iPhones on Gumtree

A first glance

Gumtree was scraped for iPhone listings using a really cool tool called Advanced Web Scraper. It makes web scraping dead-easy and I suggest you check it out. The data was then parsed from a raw JSON into an R data frame using the following function:

Now let’s take a look at the frequency density of iPhone prices:

Okay, so we see a huge peak centred at ~£150 and a secondary peak at ~£350. Why would this be the case? Perhaps there’s a glut of older models on the market and then a big jump between versions. The iPhone 6 and 6S will probably command a much higher price than the iPhone 5, for example, but would a £200 price difference be expected?

Let’s take a look at the relative numbers of the different iPhone models listed on Gumtree.

Classifying by model

To figure out what model iPhone each listing is advertising we’ll have a look at the advert title, searching for the inclusion of the iPhone model:

This piece of code cycles through the different iPhone models in the list and, if found in the listing title, the model is added to a new column. Finally, using the complete.cases() function we remove all listings that don’t specify a model. Most of these were bogus listings anyway. Now let’s plot the iPhone model frequency as a bar chart:

This perhaps isn’t too surprising: over time the value of older models decreases providing less incentive to sell, and also, phones have a habit of getting dropped and broken, futher reducing supply over time. It’s more interesting, however, that the 5S and 6 appear in equally large amounts and the most recent iPhone, the 6S, is the 3rd most commonly listed iPhone. This, despite there not being anything for iPhone users to upgrade to. Now, how much are these different models typically fetching? Let’s draw some boxplots to look at the median and spread:

Okay, wow. That was unexpected. The median price is almost constant across all models. Clearly there’s something going on here. Let’s have a closer look at the price distributions.