Diving into retail investing behaviour with Robinhood

Published in

Prosperity Advisers DnA

5 min readJun 17, 2020

Introduction

With many folks out of work or on shortened hours both in Australia and overseas there have been a number of news articles pointing out the rise in retail trading.

In Australia, we saw ASIC release an article on this (LINK) which highlighted trading activity and potential harm to retail traders jumping into the markets here (and an especially concerning piece on those jumping into CFD’s — contracts for difference).

Whilst we do not have readily available data on retail trading here in Australia, we can get some insights into markets overseas, particularly in the USA.

NB: This is not investment advice. It is general analysis and republication of existing data in an alternative user-friendly format. For advice on investments then please contact your financial adviser or get in touch with one of ours here at Prosperity: LINK

The Robinhood Data

In the USA, the mobile app Robinhood (LINK) is quite popular amongst younger traders and when you consider that the majority of jobs lost have been for people in younger age brackets (hospitality, retail, arts jobs) then insights into this app would be quite interesting.

Luckily for us, Robinhoood user tracking data is made available by RobinTrack.net (LINK). This site, run by Casey Primozic, showcases a number of interesting leaderboards which are filterable to showcase various changes in popularity of particular stocks over different periods.

Most importantly, he makes the data available to download, so that’s what I did so that I could analyse it in Power BI.

Analysis using Power BI

After downloading it, I noticed all the individual securities have their own file.

To handle this, there is a Folder analysis option when you “Get Data” in Power BI and after pulling this in we were up and rolling.

One of the things I noticed was that each security has multiple timestamps per date where user holding info is collected. To handle this we aggregate by those dates and average the number of users holding that stock on that date. This is why in the final result you’ll see the numbers of users holding a particular stock have decimal points, because we may have an average of 550.5 users who held the stock that day.

To give the app a nice template background, I create that in PPT first and load it into a blank Power BI canvas. This gives a good base for where to set your charts, KPI’s/metrics and filters.

Then I added these element using the Power BI desktop app and change display options, set colours, edit levels of interactivity and generally try to make the app simple and user-friendly.

This app allows users to search for what the biggest movers were in the 8.4k tradable securities on Robinhood over a filterable period of time. The % change takes what the total holders were at the start of the period filtered and compares that with the end number of users.

I also pulled in names of companies for 95% of the tickers listed here (from Google Finance) and also tagged those which (as at 3 June 2020) were in the S&P 500. This means users can filter for companies in the S&P 500 or not when they do their analysis.

In terms of users, I had to aggregate these and find the average users per day. This is because the data is being pulled multiple times in a day and included timestamps. So what I did there was aggregate those figures and find the average users holding the stock on a single day.

More will be added to this to do other sorts of analysis but this is it for now.

Insights

With this app we can create all sorts of insights from it. The following insights are basic but give you a good idea of what you can do with the data.

Looking at data from 1 January 2019 to 12 June 2019, we can see that users holding stocks on the platform grew +35%.

For the same time period in 2020, that growth is much more significant at +162%.

We can also see how popular the Robinhood.com domain is in terms of page views and how that’s picked up since the pandemic came into full effect. Using insights from SimilarWeb enables this (although this data is not in the app).