Introduction

With each passing day, the cost of shelter becomes increasingly expensive. Rising house prices have been discussed at length. With the median house price recently surpassing $1,000,000 NZD (Per RNZ) many Kiwi’s are being priced out of ever owning their own home. Facing a lifetime of renting ahead of them, it is worthwhile taking the time to look more closely at the trends in this ever growing market.

Methods

The New Zealand Census was used as the primary source of information for this project. It includes questions not only about the dwelling itself, but also the owners and occupiers (See here for a full breakdown). Taken every 5 years^[1] the data allows us to inspect what trends may be occurring in the data. A curated dataset was available from the New Zealand open data portal, which was downloaded and imported into Tableau.

Using Tableau to inspect the data, the trend was simple and obvious: rent is ever increasing. Of course the rate of this change does depend on a number of factors, several of which are baked into the data. Both the type of landlord and the household dynamic of the residents were mapped to time series charts, which is the tried and true method when presenting this sort of data:

Both Landlord and Household types had limited categories once filtered and grouped (4 and 6, respectively). However when attempting to plot the data broken down by region, it became somewhat unwieldily:

Thankfully, when searching for other examples of how similar data has been visualised, this example was found which provided the inspiration for how to better show this sort of data. Four charts were then created using tableau mapping module, filtered for each census. A side effect of this was that it determined at what regional level the data would be shown, as the built in map didn’t allow for finer granulation found in the data.

The four maps and two time series charts were then placed into a tableau dashboard, with the maps in sequence and the time series below. Titles were made and legends were positioned where appropriate. While the dashboard was designed to stand alone as an image, several interactive filters were added to further explore the data: The ability to select regions to filter by location (and a prompt to indicate this was possible) as well as filters for the Household and Landlord types. Click filtering was not enabled for the time series as it filtered by the exact point rather than the category. The final result being:

Discussion

Raw rent was used as the measurement, as opposed to increase e.g., as it felt as though it had the most dramatic effect. Seeing a $500 pw rent seemed more evocative than a $30 or 5% increase. It also allowed the maps to show not only increase, but where in the country rent was the most expensive (and every Kiwi knows it’s good to have a laugh at Auckland). More consideration would have been given to proportional increase if the trend wasn’t reasonably linear.

I’m quite fond of the series of heat maps, although testing is required to determine the effectiveness. It helps that New Zealand is a long country and can almost tesselate, so the effect may be diminished for a wider country such as the USA. The decision was also made to keep the heat legend consistent, as it made for easier comparison. (N.b an attempt was made to create a higher contrast colour series, but this proved difficult in Tableau). The time series charts to display prices over time was consistent with other visualisations.

Groupings and filters were created for all three charts (Regions, Households, Landlord). This excluded groups such as ‘total’, ‘other not identified’ and ‘total identified’ as they ultimately didn’t add anything to the chart. Regions used only the data for Regional Territories as it was compatible with Tableau’s map builder. Households rolled “and other person(s)” into their respective categories to limit the amount of data on screen. Landlord types excluded several categories that was only counted in the most recent census. All facets had their axes set to a fixed length to ensure a consistent comparison across filters.

Some difficulty was also had when it came to Tableau’s dashboard builder. While working within the grid system was simple, going beyond it was less so. Free floating elements were unable to be aligned or placed with accuracy, and often times would shift between the local dashboard and the published version. Dynamic resizing of the dashboard was turned off for this reason, as items would collide in smaller views. Having four maps also proved to be a challenge: Design wise they each needed to be positioned individually, and usage wise they would zoom out to a world view if a user wasn’t careful where they were scrolling.

Reflection

Overall I feel satisfied with the final dashboard, although it does have its quirks. If the project were to be revisited it would need to be done so with additional data sources, as that was a key limiting factor. Additional factors such as income, inflation, house types, and counts would all allow for a richer analysis.

Consideration would also go towards the medium of the visualisation. Tableau provides a wonderful layer of interactivity to filter the data on the dashboard, but the lack of in-depth design tools makes me less inclined to use it for static images.