# Additional datasets: Annual T1FF taxfiler data

## Background

Through collaboration with the Canadian Mortgage and Housing Corporation (CMHC), CensusMapper has added and open-sourced annual T1FF taxfiler data which provides an annual look at some basic demographic variables. Data is available via the cancensus package for the years 2001 through 2017. The T1FF dataset contains information on:

• individual income
• government transfers
• family income
• family composition
• taxfilers by age groups
• taxfilers and dependents by age groups
• marital status and select other demographic variables

The data comes in varying Census geographies, depending on the year. Retrieving any annual dataset via get_census will automatically reference to the correct Census geography and attach the correct spatial boundaries.

The taxfiler data is organized with consistent internal referencing. The identifier for the number of families in low income in 2017 is “v_TX2017_786” and that for all families is “v_TX2017_607”, and the ones for the other years are given by simply swapping out the year. This makes the variables selection process easy.

## Example usage: constructing a multi-year series of families in low-income status

As an example we will explore a multi-year time series for families in low income. Data on low income families is available for years 2004 and later, we will start with 2006 just so that the data fits on a nice grid.

# Packages used for example
library(cancensus)
library(dplyr)
library(tidyr)
library(ggplot2)
library(sf)

To see all available T1FF datasets and their reference codes we can use list_census_datasets().

list_census_datasets() %>%
filter(grepl("taxfiler",description))

And, as an example, available data vectors for one such T1FF dataset.

list_census_vectors('TX2017')

This particular dataset has over 800 individual vectors. The vector codes follow a regular pattern across different years, and we can use this to quickly identify all the relevant variables of interest across multiple datasets. We can utilized the CensusMapper graphical variable selection interface, which can also be reached by calling explore_census_vectors() from the R console. For this example we are interested in low income families and note that the internal CensusMapper vector for all families is of the form *v_TX_607* and that for all families in low income is *v_TX_786*.

While the geography varies across Census periods, the call to get_census will automatically attach the correct geography for each annual dataset. We pick four years to look at low income families.

years <- c(2006,2011,2014,2018)
# Attribution for the dataset to be used in graphs

plot_data <- years %>%
lapply(function(year) {
dataset <- paste0("TX",year)
vectors <- c("Families"=paste0("v_",dataset,"_607"),
"CFLIM-AT"=paste0("v_",dataset,"_786"))

get_census(dataset,regions=list(CMA="59933"),vectors = vectors,
geo_format = 'sf', level="CT", quiet = TRUE) %>%
select(c("GeoUID",names(vectors))) %>%
mutate(Year=year)
}) %>%
bind_rows() %>%
mutate(share=CFLIM-AT/Families)

Here we also re-organized the data by year. All that’s left is to plot the data, one year at a time.

ggplot(plot_data,aes(fill=share)) +
geom_sf(size=0.1,color="white") +
facet_wrap("Year") +
scale_fill_viridis_c(labels=scales::percent,option = "inferno",
trans="log",breaks = c(0.05,0.1,0.2,0.4)) +
coord_sf(datum=NA,xlim=c(-123.4, -122.5), ylim=c(49.01, 49.4)) +
labs(title="Share of census families in low income",fill="Share",
caption=attribution)

We may be tempted to re-arrange the data to create timelines, but we have to be careful as census geographies change over time. Inspecting the dataset tables at the top informs us that the 2006 through 2011 data all come on the common 2006 census geography, so the 2006 and 2011 tax data are directly comparable.

change_data <- plot_data %>%
filter(Year==2006) %>%
select(GeoUID,2006=share) %>%
left_join(plot_data %>%
st_set_geometry(NULL) %>%
filter(Year==2011) %>%
select(GeoUID,2011=share),
by="GeoUID") %>%
mutate(change=2011-2006)

ggplot(change_data,aes(fill=change)) +
geom_sf(size=0.1) +
labs(title="Change in share of census families in low income 2006-2011",fill="Percentage\npoint change",caption=dataset_attribution(paste0("TX",c(2006,2011))))