Pulsecheck: Is Orchestral Music Still a Living Art Form in 2017?

Extrapolating trends from 175 seasons at the New York Philharmonic

November 2017


When people hear "classical," they think "dead." The music is described in terms of its distance from the present[...] But the same story could have been written ten years ago or twenty. If this be death, the record is skipping. A complete version of the Death of Classical Music Archive would go back to the fourteenth century...-Alex Ross, "Listen to This", The New Yorker (Feb 2004)

Throughout the history of Classical Music, the perception that the genre's on its deathbed has always been a recurring motif. But how accurate are these perceptions today? What is Classical Music's current state of health? Instead of relying on anecdotal evidence, let's try to come to a conclusion using data instead.

The 2016-2017 season marked the 175th anniversary of the New York Philharmonic, and at nearly two centuries old, the oldest continually-performing American orchestra is almost four decades older than its closest sibling, the Boston Symphony Orchestra (founded in 1881). In fact, the NYP was in existence well before the births of some of the most performed and well-known Classical composers, names like Debussy, Ravel, Shostakovich, Mahler, and Prokofiev. This means that the earliest performances by the NYP stretch back deep into the Golden Age of the genre, and give us some nice benchmarks for comparison. So how does a typical season of the Philharmonic from the late 1800s compare to today? Is Classical Music really "dead" and what does the data say?

Let's dig in.

Theme: Trends from 175 Years at the New York Philharmonic

The first metric we'll explore is what I'm calling 'The Novelty Factor'. Is the orchestra exploring new pieces they've never performed before, or are they repeating music they've played over-and-over again? I plotted the number of unique compositions* performed each season by the New York Philharmonic on their Subscription Series concerts. The orange area represents the pieces being played for the first time by the NYP. In contrast, the blue area represents pieces that have been played before in a previous season. Other than the first season (1842-43), every subsequent season has featured at least some pieces that the orchestra had previously performed.

At first glance, there's not much of a pattern. The absolute number of pieces programmed each year (and in both categories) fluctuates quite unpredictably. The data is also complicated by the dramatic spike in the total number of pieces played per season, starting in the 1909-10 season. To make more sense of our data, we'll have to normalize it, and look at the ratios of 1st-time-to-repeat pieces each season instead.

Now we're comparing apples to apples. However, even this normalized data representing a ratio is quite difficult to decipher. The level of noise is still too high for us to see any obvious trends. We'll have to transform our data further: this time, we want to filter out some of the noise.

I applied a simple central moving average** to each season to smooth out the data. Beginning in the mid-1860s, the percentage of pieces being played for the first time starts to slowly, but steadily decline, from around 35% to the ~25% that is today. I was very surprised to see how early into its existence the orchestra essentially became a 'Greatest Hits' band. In less than two decades (by the 1859-60 season) of its founding, the orchestra is already, on average, repeating more pieces than they are playing pieces for the first time.

**With a sample window of 7 seasons, 3 on each side of the central season. This filter is applied on the data for all the remaining graphs in this section.

Next, let's take the 'dead or alive' metaphor literally. Is the orchestra still performing music by living composers? Compare the percentage of pieces by dead composers with those by living composers: Again, surprisingly, for most of the NYP's history (save for a few early seasons), they've performed more dead composers than living. Even when Wagner was still writing his operas, and when Brahms had yet to compose his symphonies, the Philharmonic was already mostly playing the music of 'Dead White Guys' (yes, already dead back then). Still, it's worth noting that there is a long-term decline in the percentage of performances by living composers, and this percentage collapses dramatically in the late 20th century. In the 1967-68 season, the average percentage of living composers dipped below 20% and has never recovered since. Since the 1990s, this percentage has hovered in the low teens.

Finally, let's combine our two metrics (repeat/first-time and dead/alive) into a hybrid metric. Many Billboard-charting songs get there through repeated plays on Spotify, the radio, or at clubs by DJs. The more we hear something, the more familiar it gets, and the greater the chance it'll become popular. Are living Classical composers getting this sort of repeated play? For most of the NYP's history, the share of pieces by living composers that are also repeat performances is relatively steady (around 13 to 12; look at the dotted trend line relative to the percentage of living composers). But by the 1990s, this share plummets to around 10%. As an absolute percentage, the proportion of all performances that are repeat performances of pieces by living composers today is almost negligible (percentages in the low single digits). Living composers are still getting some performances these days, but they almost never hear their pieces again.

So based on these metrics, things seem pretty dire and dicey for Classical Music, right? But it turns out, things aren't so simple. Even if lots of living composers are being performed, it doesn't mean those pieces will survive the sands of time, and just because the composers being performed are dead, doesn't mean their music can't bring vitality and novelty to the concert hall.


In this analysis, regardless of how many times a piece is performed during a single season, it is only counted once. Furthermore, if two different excerpts or movements of a piece are performed on different concerts in the same season, the piece is also treated as one single unique piece for the season. This isn't a perfect approach, since there are some scenarios in which certain 'movements' of larger works are often traditionally programmed as a standalone piece by orchestras, such as certain overtures of operas. On this front, the NYP's categorizations in their data set are sometimes inconsistent. For example, it treats the three movements of Debussy's Images as separate compositions, while each of the six pieces from Smetana's Ma Vlast are treated as 'movements'. I've preserved the orchestra's original categorizations since it's not feasible to manually recategorize all pieces.

I've attempted to exclude most non-orchestral compositions (i.e., pieces that don't require a conductor) that occasionally appear on a Subscription Series concert. For example, sometimes a soloist will perform a short solo encore piece after they play a concerto with the orchestra. Most of the known cases in this category have been omitted from the analysis and these pieces are not counted towards the number of unique compositions in a season, though I can't be 100% certain that all cases have been caught, espeically if there were miscategorizations or ambiguity in the original data (e.g., an entry that accidentally attributed a conductor in the metadata of a solo instrumental or chamber work). Arrangements of non-orchestral pieces for orchestra however, are counted.

Variations: Composers & Their Differing Destinies

Now that we've looked at the data from a high level, let's zoom in and explore what composers and music are actually being played.

How do we define a composer who's made it? There are many ways we can rank composers, but the easiest is probably to stick with our earlier method of counting unique compositions each season. A more successful composer, will simply be one who has racked up more performances by the orchestra. Though this method is flawed statistically (since it's biased in favor of composers who were alive earlier in the NYP's history, and penalizes composers born later, including those alive today), it can nevertheless give us a chance to make some other interesting observations (say the differences in career trajectories between two composers who have received around the same number of performances, or lived around the same time).

It turns out that the success of composers seems to follow a classic 'long-tail' distribution: a small subset of composers receive a disproportionate number of performances. The top-59 most performed composers (actually a total of 61, with 3 composers sharing 59th place) represent over 75% of all pieces programmed while accounting for just under 5.8% of all composers ever performed on a subscription concert (61 out of 1039).

But even within this elite group, we find a range and variety in the patterns of success (or in some cases, of decline), of how the composers' music came to be frequently performed by the New York Philharmonic:

You can explore the New York Philharmonic's performance history of these 61 composers by season below. Search for your favorite composer and check if they made the list. Do you see any other trends or patterns?

Looks like you're reading this essay on a small screen! To enjoy the full interactive experience and to explore the individual pieces of each composer performed each season, either expand your browser window or use a device with larger screen dimensions.

Each dot represents a unique composition programmed + performed in a particular season.

HOVER over the dots to see some metadata about each composition

CLICK on the dot to highlight all appearances of that piece across the NYP's history.

First and only performance of piece

First performance of a piece by the NYP

Repeat performance by the NYP

Lifespan of composer (if overlaps in part or whole with NYP history)

The 61 Most-Performed Composers by the NY Philharmonic

The 61 most performed composers by the NYP ranked by total number of performances of unique pieces per season over 175 seasons

Lifespan of composer (if overlaps in part or whole with NYP history)


Though the New York Philharmonic is just one orchestra out of many in the US, and indeed, in the world, its long, continuous existence gives us an expansive temporal-panoramic glimpse into the ebbs and flows of an artistic genre. In this essay, I primarily used two distinct dimensions in the NYP's history to pulsecheck the genre: the ratios of dead-to-living composers performed, and of repeat-vs-first-time compositions. In both measures, the NYP is clearly increasingly trending towards the repeats and the dead. But is this evidence the genre is dying?

As Alex Ross, the music critic at The New Yorker, has suggested, Classical Music has always had an eye towards the past, and from our data, we can see that even in its heyday, the New York Philharmonic was already mostly playing the works of dead composers, and happily playing those pieces over and over again.

And as we saw with composers like Spohr and Rubinstein, just because an orchestra is performing a lot of works by a living composer is no guarantee that music will last. Likewise, composers like Béla Bartók and Ives show that orchestras can discover new music and expand the repertoire with formerly-obscure composers who have long passed. 'Current' and 'now' are not the only sources of novelty.

For fans of orchestral music, there are also other reasons to be hopeful that this is still a thriving genre. I have a nagging feeling that many younger aficionados of orchestral music come to the genre not from the concert hall repertoire, but from the orchestral scores of film and video games. From John Williams to Joe Hisaishi to Michael Giacchino, the soundtracks to our modern entertainment are still rich with the sounds of innovations birthed from the concert hall, and now, musicians of all stripes are taking the centuries of innovation out of the Classical concert hall and applying them everywhere else. Maybe one day, their innovations will find their way back into the Classical concert hall. It probably already has.

Notes, Sources, & Methodology

The historical performance and concert data of the New York Philharmonic is publicly available through the orchestra's GitHub page and through their Digital Archives (Open Data). The data is very comprehensive: in addition to subscription series data, the orchestra also has concert data for the New York Symphony, founded independently in 1878, and which eventually merged with the Philharmonic in 1928. The performance data of the New York Symphony is not included in this analysis.

Beyond concerts that are part of the main Subscription Series, the Philharmonic also presents, and has presented, concerts under various other named series. As fewer and fewer contemporary compositions are featured on the main Subscription Series concerts, more and more music by living composers have been relegated to one-off and (sometimes) short-lived special festivals like Horizons (for a few seasons in the 1980s) and the more recent Biennial festival, as well as special concerts such as those that are part of the chamber new music series CONTACT! (started in the 2009-10 season). There are many factors that make it hard to compare these concerts to the main Subscription Series concerts. The venues are often much smaller than the Subscription Series concerts. In modern times, the average subscription concert program (that is, the same set of pieces) is performed multiple times, but these special concerts are often not. There's also a lot of other variation: The pieces performed on CONTACT! are also usually for smaller groups of instruments (rather than for a medium or large-sized orchestra), while some concerts as part of Horizons and the Biennial were performed by orchestras other than the New York Philharmonic.

I've attempted to correct a few mis-categorized concerts I noticed in the NYP's original JSON data, but I can't guarantee that all such mistakes are necessarily accounted for.

To calculate the number of compositions composed by living and dead composers for each season, I collected some metadata about each composer, specifically birth and death years, through Wikipedia, via a script that leveraged the MediaWiki API. A small percentage of composers had no Wiki entries, and the birth and death years for these composers, if available, were collected manually, from a variety of sources found primarily through Google. The careful reader may have notice that in the living/dead composers chart, not every season adds up to 100%. For a small number of very obscure composers, the birth and/or death years could not be verified. In addition, some seasons also featured works that were arrangements of music, such as traditional folk songs, that have no clearly attributable composer.

For readers interested in seeing or exploring the raw data that I generated from the original NYP data as part of the cleaning and analytical process and any code associated with this project (e.g., master composer list with works performed each season, script(s) for pulling composer metadata, moving average algo, D3 scripts, etc.), or who would like to file a typo or bug report, feel free to reach out to me at @ericwilliamlin on Twitter.