A quantitative approach to investing (and profiting) with music|| The Introduction ||Most consumers and investors are very familiar with the content streaming side of commercial music through services like Spotify and Apple Music and how to invest in them (simply buy $SPOT or $AAPL), but what about the production side?I (as someone who cannot go two hours without listening to good music) wanted to combine some of my interests and find out how one invests in music production and if there was a unique investment opportunity in the space that allows for a quick profit? After some thought, I narrowed my focus and formed the following research question:Are there certain stocks that are significantly affected by music output (such as songs, albums, etc.) that I can invest in? If so, can I predict which albums/songs will move the price enough to get a sizeable return ahead of the market?The answer is a resounding maybe!|| The Overview ||In the music world, there are three “overlords” of music production: Universal Music Group, Sony Music, and Warner Music Group. Within these groups are record labels, which are subsets that manage their own artists and producers and are responsible for finding promising talent/providing them with the resources necessary to succeed (and make money for the label/overarching group). Of these three, Sony Music and Warner Music Group have parent companies that gain most of their revenue from other streams (Sony Music and Warner Bros. Music encompass less than 3% of their parents’ revenues). Universal Music Group (UMG) is the exception — its parent company, an Italian media conglomerate called Vivendi, relies on UMG for ~40% of their total revenue (the other 60% of which comes from a gaming company called “Gameloft” and a sports channel called “Canal+”). This implies that any music production by UMG affects ~40% of Vivendi’s stock price, which is enough to potentially form a tradeable signal.This looked much more promising to me, and further research revealed that UMG actually owns the following record labels:XO, Interscope Records, Def Jam Recordings, Capitol Records, and Republic Records.Artists signed to these groups include The Weeknd, Eminem, Kanye West, Coldplay, Jay-Z, Rihanna, Lil Wayne, Imagine Dragons, OneRepublic, and Maroon 5 — I would never have guessed that all of these famous artists were generating revenue for the same company!My goal was then to figure out if the release of blockbuster albums by these artists (and a plethora of other UMG signees) created a positive signal that could be used to intelligently trade Vivendi’s stock.|| The Data ||I needed a way to find commonalities between the albums released by UMG artists since if every album was dissimilar with the others, any price increase is most likely independent of the albums themselves.I settled on using album certifications (silver, gold, platinum, etc.), first-week sales, number of tracks, album type (single, compilation, etc.), and release date as potential factors that could be used to find similarities. I could now start trying to pull these factors together into a dataset.Vivendi’s Stock PriceI used Yahoo Finance to extract daily close prices for “Vivendi SA” ($VIV.PA), but this did not suit my purpose since I wanted to measure the effect of an album release over a specific time window. I chose a window of seven trading days after the album release (because first-week sales is a common metric in the music industry) and calculated rolling percent changes.Album ReleasesI scraped Genius.com’s “album release schedule” to get a list of album drops and their respective artists for the last two years (2016, 2017).Album InformationUsing the Spotify Web API, I searched for each of the albums I had found in the previous step and was able to get a little more information on each of them — specifically, I was able to extract the number of tracks on the album, the album type (single, compilation, etc.), and their specific release dates.UMG ArtistsI also scraped a list of UMG artists from a Wikipedia article and cross-referenced them with the data from the previous step to get a list of 118 UMG album drops along with their respective artists and the factors mentioned above.My initial plan was to scrape certification data from the Recording Industry Association of America (RIAA), which has awards data on every album (if it received an award). Unfortunately, I made too many HTTP requests and was locked out of the website for security reasons, so I had to pivot.|| The Analysis ||The lack of album certification data forced me to resort to the binary question of whether or not the album release increased Vivendi’s stock price by greater than 3% (an arbitrarily selected number that seemed statistically significant). This pivot meant I had no factors to try and predict how much the stock price changed, only whether it changed at all.I ran the dates of the 118 albums through an algorithm that marked the ones that significantly impacted Vivendi with a target value of 1 (and the rest with a 0).At the end of this algorithm, I found that there were 25 UMG albums out of the 118 in the past two years that may have inflated Vivendi’s stock price by at least 3% during the week after its release.Factor data for all 118 album releases by UMG artists (includes target values of “0” for no significant impact and “1” for significant impact)In order to establish dependence between a 3% increase in $VIV.PA and these albums, there need to be some commonalities between the albums marked as “1.” From the charts above, it’s pretty easy to say the influential albums had little in common with each other (at least for the factors I was able to get). The last chart is the most interesting, however, because it suggests that albums with between 10–15 tracks are most likely to cause a price increase. The other charts show that albums perform best when they are released near the beginning or the end of the year and that “albums” perform much better than “singles” or “compilations” (but this may also be because most of the 118 were albums).|| The Conclusion ||In one sentence, my little experiment yielded the following conclusion:There might be evidence of a tradeable signal on Vivendi’s stock price a week after a given album release, but further work is necessary to determine how that signal can be determined or what it even is.There were also some lurking variables that may have skewed my experiment, such as the performance of the other subdivisions of Vivendi (like Gameloft and Canal+). In addition, I definitely had a limited dataset, since Genius.com only started their album release calendar in 2016.25 of 118 album drops might be enough to establish an inkling of a correlation between $VIV.PA and UMG’s album releases, but that correlation depends on the 25 albums that marked a >3% increase. If those 25 have some element in common, it becomes possible to distinguish which kind of albums cause such a spike in $VIV.PA.Unfortunately, with the limited data I was able to extract, I could not identify more telling commonalities between the albums that significantly impacted Vivendi’s stock price. Therefore, though there is a slight evidence that a correlation might exist, I cannot make any strong claims about how powerful that correlation is.So, like most things in statistics,the true answer to this problem sings an unknown tune.|| The Afterword ||I conducted this experiment with R and used a couple of packages including “ggplot2” (data visualization), “rvest” (web scraping), and “httr” (HTTP requests).I invite you to conduct your own experiments relating to this topic and would love to hear any findings or progress! Give this article some claps if you liked it and stay “tuned” for more data-driven investment idea exploration by following TDS Team and me (Abhinav Raghunathan)!