---
title: "readrba"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{readrba}
  %\VignetteEncoding{UTF-8}
  %\VignetteEngine{knitr::rmarkdown}
editor_options: 
  chunk_output_type: console
---

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.width = 5.6, 
  fig.height = 4, 
  fig.align = "center",
  fig.retina = 2,
  out.width = "100%",
  out.extra = 'style="border:0px;"'
)
```

The Reserve Bank of Australia publishes lots of useful [statistical tables on its website](https://rba.gov.au/statistics/tables/). The `{readrba}` package helps you:

1.  Download RBA statistical tables; and
2.  Import the data into R as a tidy tibble.

You can import the data using either the table number or the series ID.

```{r setup, results=FALSE, message=FALSE, warning=FALSE}
library(readrba)
library(tidyr)
library(dplyr)
library(ggplot2)
library(lubridate)

theme_set(theme_minimal())
```

### Example: Individual series - 10-year bond yield

Let's start off by visualising the yield on a 10 year Australian Government bond over time.

I want to find the series ID for the 10 year Australian Government bond yield. The function `browse_rba_series()` will help me find it. If you use the function with no argument, it will return a tibble listing every data series published by the RBA (and readable by {readrba}). I'll specify a search string:

```{r 10yearnorun, eval=FALSE}
browse_rba_series("Australian government 10 year")
```
```{r 10year, echo=FALSE}
browse_rba_series("Australian government 10 year") %>%
  knitr::kable()
```

OK, so we have two unique series for 10 year bond yield - one is daily data, the other monthly. These two series are split across four tables - a 'current' and 'historical' table for each of the daily and monthly series.

I'll get the monthly data, which we can see from the table above has the series ID `FCMYGBAG10`.

If you prefer, you can find the series ID by looking at the relevant spreadsheet from the RBA's website.

Now we'll use that series ID to download the data:
```{r bondyield, eval = F}
bond_yield <- read_rba(series_id = "FCMYGBAG10")
```
```{r, include = F, eval = F}
bond_yield <- read_rba(series_id = "FCMYGBAG10")

save(bond_yield, file = file.path("vignettes", "FCMYGBAG10.rda"))

```

```{r, include = F}
load("FCMYGBAG10.rda")
```

A quick aside: the code above gives identical results to using `read_rba_seriesid("FCMYGBAG10")`. The `read_rba_seriesid("SOMETHING")` function is a wrapper around `read_rba(series_id = "SOMETHING")`.

OK, now we have the bond yield data in tibble form, let's see what it looks like:

```{r glimpsebondyield}
glimpse(bond_yield)
```

The data you import using `read_rba()` is always in this standard tidy format. The values are in the `value` column.

Let's make a graph of the 10 year bond yield over time:

```{r vizbondyield}
bond_yield %>%
  ggplot(aes(x = date, y = value)) +
  geom_line()

```

That's a bit normcore-basic ggplot, but not bad. With a few lines of code, we've got an up-to-date graph of the 10 year bond yield.

### Example: Multiple series from a table - 2, 3, 5 and 10 year bond yields

Now we'll have a look at another example, this time using multiple data series from the one table. I want to make a graph of Australian Government bond yields at different maturities, over time. This time I'll confine myself to the recent data, at daily frequency.

```{r yieldbrowse, eval=FALSE}
browse_rba_tables("government bond")
```

```{r yieldbrowse2, echo=FALSE}
browse_rba_tables("government bond") %>%
  knitr::kable()
```

The daily data is in table F2. If you prefer, you can look up the table on the [RBA website](https://www.rba.gov.au/statistics/tables/) and find the table number that way.

Now we know the table number that contains the data we want, let's load it into r.

```{r variousyields, eval = F}
f2 <- read_rba("f2")
```
```{r, include = F}
load("f2.rda")
```


This table looks just like the single series table we downloaded earlier: `r glimpse(f2)`. Unlike the single series table - which only contained the ten year bond yield - we now have a tidy tibble containing a number of data series. Here's what we've got:

```{r uniquef2}
unique(f2$series)
```

I'm not interested in NSW bonds, nor indexed bonds, so we'll need some brief {dplyr}-ing of the data before we can get down to graphing.

```{r filterf2}
filtered_f2 <- f2 %>%
  filter(grepl("Australian Government", series) &
           !grepl("Indexed", series))
```

We've dropped the series we're not interested in, going from `r nrow(f2)` rows to `r nrow(filtered_f2)` rows. Now let's graph:

```{r vizf2}
filtered_f2 %>%
  ggplot(aes(x = date, y = value, col = series)) +
  geom_line()

```

Nice! Again, normcore aesthetics aside, we've got a decent plot with a few lines of code. Using the same data, we can have a look at the spread between the yield on 10 year and 2 year bonds, as a rough measure of the slope of (a portion of) the yield curve.

```{r}
filtered_f2 %>%
  select(date, series, value) %>%
  spread(key = series, value = value) %>%
  mutate(spread_10_2 = `Australian Government 10 year bond` - `Australian Government 2 year bond`) %>%
  ggplot(aes(x = date, y = spread_10_2)) +
  geom_line()
```

### Example: RBA forecasts
The package also includes all public forecasts of key economic variables made 
by the Reserve Bank since 1990. The `rba_forecasts()` function returns 
these forecasts in a tidy tibble -- including the latest forecasts, scraped
from the Statement on Monetary Policy.

```{r, eval = F}
forecasts <- rba_forecasts()
```
```{r, include = F}
forecasts <- rba_forecasts(refresh = FALSE)
```

Here's what they look like:

```{r, echo = F}
dplyr::glimpse(forecasts)
```

Let's visualise all the Bank's unemployment rate forecasts since 1990:

```{r viz-unemp-forecasts}
forecasts %>%
  filter(series == "unemp_rate") %>%
  ggplot(aes(x = date, 
             y = value, 
             group = forecast_date, 
             colour = forecast_date)) +
  geom_line()
```

### Example: change in RBA forecasts

Now let's make a chart that compares the RBA's latest forecasts to its previous forecasts.  

```{r change-in-forecasts,   fig.height = 6}
# We've already created the `forecasts` object, like this: 
# forecasts <- rba_forecasts()

latest_two <- forecasts %>%
  filter(forecast_date %in% c(max(forecast_date),
                              max(forecast_date) - months(3)))
latest_two %>% 
  group_by(series_desc) %>% 
  # Only include series for which we have forecasts in both of the latest SMPs
  filter(length(unique(forecast_date)) >= 2) %>% 
  ungroup() %>% 
  mutate(forecast_date = format(forecast_date, "%b %Y")) %>%
  ggplot(aes(x = date, y = value, col = forecast_date)) +
  geom_line() +
  guides(colour = guide_legend(title = "Forecast issued: ")) +
  facet_wrap(~series_desc, scales = "free_y",
             labeller = label_wrap_gen(17)) +
  theme_minimal(base_size = 12) +
  theme(legend.position = "top",
        legend.direction = "horizontal",
        legend.title = element_text(),
        axis.title = element_blank(),
        axis.text = element_text(size = 6),
        plot.margin = margin(),
        strip.text = element_text(size = 8)) +
  labs(subtitle = paste0("RBA's forecasts issued in ",
                         unique(latest_two$forecast_date) %>%  format("%B %Y") %>%  paste(collapse = " and ")),
       caption = "Source: RBA Statement on Monetary Policy")

```