Using readabs

Matt Cowgill

When working with time series data from the Australian Bureau of Statistics (ABS), you must:

  1. Download the data;
  2. Read the data into R; and
  3. Tidy the data.

The readabs package provides functions to help you with each of those steps. One key function will help streamline the process of analysing ABS time series data:

The separate_series() function also helps with step three, tidying the data. You use it in conjunction with read_abs() - see the examples below.

Another function, read_abs_local() is useful if you have already downloaded ABS time series spreadsheet to disk; it imports and tidies the spreadsheets.

The messiness of ABS time series

If you want to visualise or analyse data in R, you will often need to tidy it first. In tidy data:

  1. Each variable forms a column.
  2. Each observation forms a row.
  3. Each type of observational unit forms a table.

ABS time series data is not tidy. Tidying it requires a bit of work. This screenshot of an ABS time series spreadsheet shows some of the problems, namely:

readabs does a lot of the work of tidying these spreadsheets for you, so you can get to your analysis more quickly.

readabs only works with time series

The spreadsheets on the ABS website are divided into one of two categories: time series spreadsheets and data cubes. For example, the main Labour Force release contains both:

The readabs package can download and tidy data contained in ABS time series spreadsheets. It can’t download or tidy any spreadsheet the ABS describes as a ‘data cube’.

How to use read_abs() to get a whole catalogue number

The main function in the package is read_abs(). If you give it an ABS catalogue number, it will download, import and tidy all the time series spreadsheets from that catalogue number. Easy!

For example, to get all the spreadsheets from the Wage Price Index, catalogue number 6345.0, we’d do:

library(readabs)

wpi <- read_abs("6345.0")
#> Finding filenames for tables corresponding to ABS catalogue 6345.0
#> Attempting to download files from catalogue 6345.0, Wage Price Index, Australia
#> Extracting data from downloaded spreadsheets
#> Tidying data from imported ABS spreadsheets

Cool! Now we’ve got a data frame (a tibble, to be precise) that contains all the time series from the Wage Price Index, converted to long and stacked on top of each other. Here’s what it looks like:

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union

glimpse(wpi)
#> Observations: 56,276
#> Variables: 12
#> $ table_no         <chr> "634501", "634501", "634501", "634501", "634501…
#> $ sheet_no         <chr> "Data1", "Data1", "Data1", "Data1", "Data1", "D…
#> $ table_title      <chr> "Table 1. Total Hourly Rates of Pay Excluding B…
#> $ date             <date> 1997-09-01, 1997-12-01, 1998-03-01, 1998-06-01…
#> $ series           <chr> "Quarterly Index ;  Total hourly rates of pay e…
#> $ value            <dbl> 67.4, 67.9, 68.5, 68.8, 69.6, 70.0, 70.4, 70.8,…
#> $ series_type      <chr> "Original", "Original", "Original", "Original",…
#> $ data_type        <chr> "INDEX", "INDEX", "INDEX", "INDEX", "INDEX", "I…
#> $ collection_month <chr> "3", "3", "3", "3", "3", "3", "3", "3", "3", "3…
#> $ frequency        <chr> "Quarter", "Quarter", "Quarter", "Quarter", "Qu…
#> $ series_id        <chr> "A2603039T", "A2603039T", "A2603039T", "A260303…
#> $ unit             <chr> "Index Numbers", "Index Numbers", "Index Number…

It’s over 54 000 rows long, and 12 variables wide. Some catalogue numbers are much bigger - for example, if you get the entire monthly Labour Force release (catalogue number 6202.0), you’ll have a data frame with over 2.1 million rows.

All the metadata from the time series spreadsheets is included in the data frame:

The table_no and sheet_no columns will help you if you need to cross-check information on the ABS spreadsheet - table_no matches the filename of the spreadsheet (eg. ‘634501.xls’) and sheet_no is the name of the Excel worksheet within the file that contains the time series.

How to use read_abs() to get individual table(s)

Unless you tell it otherwise, read_abs() will get all the time series spreadsheets from a given catalogue number.

Generally this will be overkill. Maybe you don’t want all 2.1 million rows of Labour Force data; perhaps you know that the time series you need is in table 1. In that case you can use the tables argument to read_abs() to specify the table(s) you want:


lfs_1 <- read_abs("6202.0", tables = 1)
#> Finding filenames for tables corresponding to ABS catalogue 6202.0
#> Attempting to download files from catalogue 6202.0, Labour Force, Australia
#> Extracting data from downloaded spreadsheets
#> Tidying data from imported ABS spreadsheets

glimpse(lfs_1)
#> Observations: 56,772
#> Variables: 12
#> $ table_no         <chr> "6202001", "6202001", "6202001", "6202001", "62…
#> $ sheet_no         <chr> "Data1", "Data1", "Data1", "Data1", "Data1", "D…
#> $ table_title      <chr> "Table 1. Labour force status by Sex, Australia…
#> $ date             <date> 1978-02-01, 1978-03-01, 1978-04-01, 1978-05-01…
#> $ series           <chr> "Employed total ;  Persons ;", "Employed total …
#> $ value            <dbl> 6008.280, 6015.217, 6021.606, 6027.255, 6031.25…
#> $ series_type      <chr> "Trend", "Trend", "Trend", "Trend", "Trend", "T…
#> $ data_type        <chr> "STOCK", "STOCK", "STOCK", "STOCK", "STOCK", "S…
#> $ collection_month <chr> "1", "1", "1", "1", "1", "1", "1", "1", "1", "1…
#> $ frequency        <chr> "Month", "Month", "Month", "Month", "Month", "M…
#> $ series_id        <chr> "A84423127L", "A84423127L", "A84423127L", "A844…
#> $ unit             <chr> "000", "000", "000", "000", "000", "000", "000"…

If you want more than one table, but not the whole catalogue number, you can specify multiple tables:


lfs_1_5 <- read_abs("6202.0", tables = c(1, 5))
#> Finding filenames for tables corresponding to ABS catalogue 6202.0
#> Attempting to download files from catalogue 6202.0, Labour Force, Australia
#> Extracting data from downloaded spreadsheets
#> Tidying data from imported ABS spreadsheets

glimpse(lfs_1_5)
#> Observations: 98,604
#> Variables: 12
#> $ table_no         <chr> "6202001", "6202001", "6202001", "6202001", "62…
#> $ sheet_no         <chr> "Data1", "Data1", "Data1", "Data1", "Data1", "D…
#> $ table_title      <chr> "Table 1. Labour force status by Sex, Australia…
#> $ date             <date> 1978-02-01, 1978-03-01, 1978-04-01, 1978-05-01…
#> $ series           <chr> "Employed total ;  Persons ;", "Employed total …
#> $ value            <dbl> 6008.280, 6015.217, 6021.606, 6027.255, 6031.25…
#> $ series_type      <chr> "Trend", "Trend", "Trend", "Trend", "Trend", "T…
#> $ data_type        <chr> "STOCK", "STOCK", "STOCK", "STOCK", "STOCK", "S…
#> $ collection_month <chr> "1", "1", "1", "1", "1", "1", "1", "1", "1", "1…
#> $ frequency        <chr> "Month", "Month", "Month", "Month", "Month", "M…
#> $ series_id        <chr> "A84423127L", "A84423127L", "A84423127L", "A844…
#> $ unit             <chr> "000", "000", "000", "000", "000", "000", "000"…

The tables argument can be either a numeric vector (eg. c(1, 5)) or a character vector (eg. c(“1”, “5a”)).

I’ve imported the data… now what?

Because read_abs() does the work for you of getting your data in a tidy (long) format, it’s easier to filter to the data you’re interested in and generate output (like graphs) using the Tidyverse packages, such as dplyr and ggplot2.

In this example, we’ll work with data from the Labour Force survey that we downloaded and tidied earlier using read_abs(). First, load the packages you need:

Now let’s have a look at the time series from table 1 of the Labour Force survey:

unique(lfs_1$series)
#>  [1] "Employed total ;  Persons ;"                                      
#>  [2] "Employed total ;  > Males ;"                                      
#>  [3] "Employed total ;  > Females ;"                                    
#>  [4] "> Employed full-time ;  Persons ;"                                
#>  [5] "> Employed full-time ;  > Males ;"                                
#>  [6] "> Employed full-time ;  > Females ;"                              
#>  [7] "> Employed part-time ;  Persons ;"                                
#>  [8] "> Employed part-time ;  > Males ;"                                
#>  [9] "> Employed part-time ;  > Females ;"                              
#> [10] "Employment to population ratio ;  Persons ;"                      
#> [11] "Employment to population ratio ;  > Males ;"                      
#> [12] "Employment to population ratio ;  > Females ;"                    
#> [13] "Unemployed total ;  Persons ;"                                    
#> [14] "Unemployed total ;  > Males ;"                                    
#> [15] "Unemployed total ;  > Females ;"                                  
#> [16] "> Unemployed looked for full-time work ;  Persons ;"              
#> [17] "> Unemployed looked for full-time work ;  > Males ;"              
#> [18] "> Unemployed looked for full-time work ;  > Females ;"            
#> [19] "> Unemployed looked for only part-time work ;  Persons ;"         
#> [20] "> Unemployed looked for only part-time work ;  > Males ;"         
#> [21] "> Unemployed looked for only part-time work ;  > Females ;"       
#> [22] "Unemployment rate ;  Persons ;"                                   
#> [23] "Unemployment rate ;  > Males ;"                                   
#> [24] "Unemployment rate ;  > Females ;"                                 
#> [25] "> Unemployment rate looked for full-time work ;  Persons ;"       
#> [26] "> Unemployment rate looked for full-time work ;  > Males ;"       
#> [27] "> Unemployment rate looked for full-time work ;  > Females ;"     
#> [28] "> Unemployment rate looked for only part-time work ;  Persons ;"  
#> [29] "> Unemployment rate looked for only part-time work ;  > Males ;"  
#> [30] "> Unemployment rate looked for only part-time work ;  > Females ;"
#> [31] "Labour force total ;  Persons ;"                                  
#> [32] "Labour force total ;  > Males ;"                                  
#> [33] "Labour force total ;  > Females ;"                                
#> [34] "Participation rate ;  Persons ;"                                  
#> [35] "Participation rate ;  > Males ;"                                  
#> [36] "Participation rate ;  > Females ;"                                
#> [37] "Not in the labour force (NILF) ;  Persons ;"                      
#> [38] "Not in the labour force (NILF) ;  > Males ;"                      
#> [39] "Not in the labour force (NILF) ;  > Females ;"                    
#> [40] "Civilian population aged 15 years and over ;  Persons ;"          
#> [41] "Civilian population aged 15 years and over ;  > Males ;"          
#> [42] "Civilian population aged 15 years and over ;  > Females ;"

OK! There’s a bunch of data in here. We can see that the series column contains a first part, like “Participation rate” or “Unemployment rate”, and a second part, which is “Persons”, “Males”, or “Females”. You will often want to filter using one or both of these components separately - and sometimes there will be more than two components.

The separate_series function helps you by separating the series column into multiple components and removing extraneous symbols.

lfs_1_sep <- lfs_1 %>% 
  separate_series()

lfs_1_sep
#> # A tibble: 56,772 x 14
#>    table_no sheet_no table_title date       series series_1 series_2 value
#>    <chr>    <chr>    <chr>       <date>     <chr>  <chr>    <chr>    <dbl>
#>  1 6202001  Data1    Table 1. L… 1978-02-01 Emplo… Employe… Persons  6008.
#>  2 6202001  Data1    Table 1. L… 1978-03-01 Emplo… Employe… Persons  6015.
#>  3 6202001  Data1    Table 1. L… 1978-04-01 Emplo… Employe… Persons  6022.
#>  4 6202001  Data1    Table 1. L… 1978-05-01 Emplo… Employe… Persons  6027.
#>  5 6202001  Data1    Table 1. L… 1978-06-01 Emplo… Employe… Persons  6031.
#>  6 6202001  Data1    Table 1. L… 1978-07-01 Emplo… Employe… Persons  6033.
#>  7 6202001  Data1    Table 1. L… 1978-08-01 Emplo… Employe… Persons  6033.
#>  8 6202001  Data1    Table 1. L… 1978-09-01 Emplo… Employe… Persons  6033.
#>  9 6202001  Data1    Table 1. L… 1978-10-01 Emplo… Employe… Persons  6035.
#> 10 6202001  Data1    Table 1. L… 1978-11-01 Emplo… Employe… Persons  6040.
#> # … with 56,762 more rows, and 6 more variables: series_type <chr>,
#> #   data_type <chr>, collection_month <chr>, frequency <chr>,
#> #   series_id <chr>, unit <chr>

The remove_totals and remove_nas arguments to separate_series() can be very useful when you are tidying data. They’re both set to FALSE by default.

We’ve now got a new tibble that still contains the original series column, but also contains new columns series_1 and series_2, with the two parts of series split into these new columns.

Let’s make a data frame that just contains the male and female unemployment rates over time, using the seasonally adjusted time series.


unemp <- lfs_1_sep %>%
  filter(series_1 == "Unemployment rate")

unique(unemp$series_1)
#> [1] "Unemployment rate"

unique(unemp$series_2)
#> [1] "Persons" "Males"   "Females"

Now we have a data frame, unemp, that contains various unemployment rate series. Let’s filter to only Males or Females.


unemp <- unemp %>%
  filter(series_2 %in% c("Males", "Females"))

unique(unemp$series_2)
#> [1] "Males"   "Females"

Now our data frame only contains the male and female unemployment rates, which is what we want. Let’s graph it, filtering once more to show only the seasonally adjusted series and adding a ‘sex’ column:

library(ggplot2)

unemp %>%
  filter(series_type == "Seasonally Adjusted") %>%
  mutate(sex = series_2) %>%
  ggplot(aes(x = date, y = value, col = sex)) +
  geom_line() +
  theme_minimal() +
  theme(legend.position = "bottom",
        axis.title = element_blank(),
        legend.title = element_blank(),
        text = element_text(size = 5)) +
  labs(title = "The male and female unemployment rates have converged",
       subtitle = "Unemployment rates for Australian men and women (aged 15+), 1978-2018 (per cent)",
       caption = "Source: ABS 6202.0")

Ta-da! Now we’ve got a nice little ggplot2 graph - and you didn’t need to go to the ABS website or click around in Excel.

How to get time series using their unique ID(s)

Every ABS time series has a unique identifier. If you know the ID for the time series you need, you can supply it to the series_id argument to read_abs(). You can give it a single ID, or multiple IDs.

For example, the time series ID for the number of employed people in Australia (trend) is “A84423127L”. We can get a data frame that only contains that series as follows:


employed <- read_abs(series_id = "A84423127L")
#> Finding filenames for tables corresponding to ABS series ID
#> Attempting to download files from series ID , Labour Force, Australia
#> Extracting data from downloaded spreadsheets
#> Tidying data from imported ABS spreadsheets

glimpse(employed)
#> Observations: 498
#> Variables: 12
#> $ table_no         <chr> "6202001", "6202001", "6202001", "6202001", "62…
#> $ sheet_no         <chr> "Data1", "Data1", "Data1", "Data1", "Data1", "D…
#> $ table_title      <chr> "Table 1. Labour force status by Sex, Australia…
#> $ date             <date> 1978-02-01, 1978-03-01, 1978-04-01, 1978-05-01…
#> $ series           <chr> "Employed total ;  Persons ;", "Employed total …
#> $ value            <dbl> 6008.280, 6015.217, 6021.606, 6027.255, 6031.25…
#> $ series_type      <chr> "Trend", "Trend", "Trend", "Trend", "Trend", "T…
#> $ data_type        <chr> "STOCK", "STOCK", "STOCK", "STOCK", "STOCK", "S…
#> $ collection_month <chr> "1", "1", "1", "1", "1", "1", "1", "1", "1", "1…
#> $ frequency        <chr> "Month", "Month", "Month", "Month", "Month", "M…
#> $ series_id        <chr> "A84423127L", "A84423127L", "A84423127L", "A844…
#> $ unit             <chr> "000", "000", "000", "000", "000", "000", "000"…

unique(employed$series)
#> [1] "Employed total ;  Persons ;"

A benefit of this approach is that the filtering is done for you - the data frame only contains the series you’re interested in.

Why is my hard drive filling up with spreadsheets?

The read_abs() function downloads spreadsheets from the ABS website, then loads them into R, then tidies them. By default, the spreadsheets will be saved in a data/ABS subdirectory of your working directory. You can change this location using the path argument to read_abs().

Keeping the spreadsheets is often a good idea. But if you don’t want to retain them, no problem. Just specify retain_files = FALSE when you use read_abs(). The spreadsheets will still be downloaded, but they’ll be stored in a temporary directory that is flushed when your R session ends.

What if I’ve already downloaded spreadsheets?

If you already have ABS time series spreadsheets saved locally that you want to read, the read_abs_local() function is what you want.

If you don’t just run read_abs_local() without any arguments, it will look in the data/ABS subdirectory of your working directory and attempt to read any .xls files located there (note: it won’t look in subdirectories of path). If you want to read all the files from a different directory, specify it using the path argument.

If you’ve downloaded files using read_abs(), you can import them using read_abs_local() by specifying the catalogue number. This will look in the subdirectory of path that corresponds to cat_no - in this case it will read all files from “data/ABS/6202.0”:

lfs_local_1 <- read_abs_local("6202.0")
#> Extracting data from locally-saved spreadsheets
#> Tidying data from imported ABS spreadsheets

If you want to read a particular table, or tables, you can specify them using the filenames argument. If your files are not in “data/ABS”, you’ll need to specify a subdirectory like this:


lfs_local_2 <- read_abs_local(filenames = c("6202001.xls", "6202005.xls"),
                            path = "data/ABS/6202.0")
#> Extracting data from locally-saved spreadsheets
#> Tidying data from imported ABS spreadsheets

The data frame you’ll get will look the same as if you’d used read_abs() to get the spreadsheet(s) from the ABS website.