IMDb · 1990–2018 · 2,266 drama seasons

The Golden Age of TV is real — but it is the floor that rose, not the ceiling.

In November 2018, The Economist’s data team plotted every American drama season ever rated on IMDb. The cloud they got is the same one you are about to see.

scroll ↓

Look at the cloud first

In November 2018, The Economist’s data team built one chart to answer one question: is TV’s Golden Age real? They pulled IMDb ratings for every American TV drama season from 1990 to October 2018 — 2,266 seasons of 868 shows — and dropped each one as a single dot on a graph of premiere date against rating. The cloud they got is the same one you are about to see.

Stare at it long enough and a famous show will pop. Game of Thrones at 9.49. Breaking Bad at 9.55. Mad Men climbing season by season. But also: Parenthood (the 1990 NBC drama, not the 2010 reboot) sitting at the top with a 9.68. L.A. Law season four at 9.40, the same as Greek season four in 2011. The very highest-rated dot in the entire dataset is from 1990.

So before we get to the answer, look at the cloud and decide what you think. Has television actually gotten better since the day The Sopranos premiered? Or have we been telling ourselves a flattering story?

Before you look — pick one
The median 1990s drama scored 7.88. The median 2010s drama scored 8.17. That is a real lift — and it is also less than a third of a point on a 10-point scale. Now look at the cloud.

A quarter of a point

The median IMDb rating for an American drama in the 1990s was 7.88. The median in the 2010s was 8.17. That is a real, sustained shift — and it is also less than a third of a point on a 10-point scale.

Split into three eras — the pre-Sopranos years (1990–1998), the cable-prestige era (1999–2012), and the streaming era (2013–2018) — and the climb shows up neatly: 7.88 → 8.07 → 8.19. Each step is bigger than the last in absolute terms, but the total lift is still only +0.31 points across nearly thirty years of television history.

That is what The Sopranos, Mad Men, Breaking Bad, and Game of Thrones added to the average drama, in the cold currency of IMDb stars: roughly a quarter of a point.

7.88
Pre-Sopranos
1990–1998
8.07
Cable prestige
1999–2012
8.19
Streaming era
2013–2018
+0.31 across nearly thirty years — the entire arithmetic of the “Golden Age”
Hear the slow lift — warm piano rises in pitch with each era’s median.

The ceiling did not move

Now look only at each year’s top 10% of dramas — the ceiling. In 1995 the 90th-percentile season scored 8.96. In 2017 the 90th-percentile season scored 8.90. Those are statistically the same number. The very best 1990s drama was as good as the very best 2010s drama.

What rose was the floor. The 10th-percentile season climbed from 7.01 in 1995 to 7.23 in 2017. The median crept up by a third of a point. The ceiling did not budge.

The top-ten list cinches it. Six of the ten highest-rated seasons in the entire dataset predate 2008 — Parenthood (1990), Homicide: Life on the Street (1997), Touched by an Angel (1998), Are You Afraid of the Dark? (1994), L.A. Law (1990), Greek (2011). The other four — Breaking Bad, two seasons of Game of Thrones, and BoJack Horseman — are real, and they are great. But there is no era in which the all-time top-ten list belongs to one decade.

The Golden Age, properly described, is not a new ceiling. It is a higher standard for what counts as forgettable.

Ten times as many shows

In 1990, American TV produced 22 drama seasons that anyone bothered to put on IMDb. In 2017 it produced 230 — a 10.5x explosion. The growth curve is monotonic from 2002 onward, only briefly nicked by the 2008 writers’ strike. By the time this dataset was assembled, an average viewer was choosing between five times more shows per year than their parents did.

John Landgraf, the FX boss who has tracked the industry’s scripted-series count for two decades, put 2017’s number at 487 across all genres. He named it Peak TV — and by 2018 was already calling it a “Gilded Age,” worried that quality could not keep pace with quantity. The data here suggests he was almost right: the median rose, but very modestly. What moved by ten times was the count.

What rising actually looks like

If the typical drama only nudged up a quarter-point, the most celebrated dramas did much more. Breaking Bad opened at 8.71 in 2008 and ended at 9.55 — a season-by-season climb so clean it could be a textbook diagram. Mad Men went from 8.23 to 8.85 over seven seasons. The Sopranos held the high 8s for six years. The Wire — famously slow, famously dense — barely moved at 8.24 to 8.42, the kind of flat line a critic would call mastery and an algorithm would call boring.

Game of Thrones is the most volatile. Open at 9.13 in 2011, peak at 9.49 in season six, dip to 9.30 in season seven. The dataset cuts off in October 2018, just before the show’s eighth and final season aired in 2019 and the IMDb score for the series finale collapsed to 4.0. Look at the season-seven dip in this data and you can already see the joints loosening.

The HBO trailer for Game of Thrones season 8 — the canonical streaming-era flame-out. The IMDb finale rating that followed (4.0/10) sits one year outside this dataset but is foreshadowed by the S7 dip you can see in the chart above. Watch on YouTube →

When the climb reverses

Dexter is the textbook prestige-era flame-out. Seven seasons in the 8.4–9.1 band — solid territory. Then in 2013, the eighth and final season scored 6.95. The character Dexter Morgan ended the show by faking his own death and reinventing himself as a Pacific Northwest lumberjack. The IMDb score fell 2.20 points from the show’s S4 peak.

It was not the worst within-show fall in the data. NYPD Blue holds that distinction with a 3.62-point drop from S1 (8.32) to S4 (4.70) in 1997. Bloodline (-1.99 between S2 and S3 in 2017), Lethal Weapon (-2.38 between S2 and S3 in 2018), and the long-running Law & Order: SVU (-2.35 between S6 and S16) are the streaming-era counterparts. Game of Thrones S8, the most famous flame-out of all, would happen one year after this dataset closes.

These collapses are partly real and partly an IMDb voting artefact: angry viewers self-select to vote a show down. The same self-selection that lifts the median above 8 in the 1990s helps push a finale into the 4s. Both directions are real and both are amplified.

Horror won, Sci-Fi lost

Empty haunted suburban living room at night with a single old CRT television glowing pale blue in the dark
Horror went from a fringe genre in the 1990s (median 7.48) to a prestige one in the 2010s (8.38). The biggest +0.90 lift in the data.

If you split dramas by genre, the Golden Age belongs to Horror. The median rating for horror-tagged dramas climbed from 7.48 in the 1990s to 8.38 in the 2010s — a 0.90-point lift, three times the all-drama average. The Walking Dead, American Horror Story, Hannibal, and Penny Dreadful turned what used to be a fringe category into a prestige category.

Adventure (+0.57) and Action (+0.45) also climbed substantially. Drama, Crime, and Romance all moved a more typical +0.29. Two genres actually declined: Thriller fell -0.30 and Sci-Fi fell -0.20. Sci-Fi is the more interesting case — Star Trek: Deep Space Nine, Babylon 5, and The X-Files set such a high bar in the early 1990s that 2010s science fiction could not keep pace by IMDb’s measure.

Three things at once

Three things have happened to American TV drama since 1990, and they are not the same thing.

10.5×
More drama seasons
in 2017 than 1990
+0.31
Median lift
pre-Sopranos → streaming
8.96 → 8.90
Ceiling 1995 → 2017
essentially flat

First, the count of shows roughly tripled by 2010 and grew tenfold by 2017. Whatever you watch in any given week, there are nine other shows you could be watching instead.

Second, the median quality rose modestly — about a quarter of a point on IMDb’s ten-point scale. The floor moved up. The ceiling did not.

Third — and this is what most criticism leaves out — the 2010s produced more good shows in absolute terms than any prior decade. Not because each one is rated higher, but because there are so many more of them. A 50% larger middle of an 8.0+ distribution that has tripled in size means several times more 8.5+ shows than the 1990s ever produced.

So when someone calls Breaking Bad a Golden Age show, they are right twice and wrong once. Breaking Bad’s S5 (9.55) is not appreciably better than Parenthood S1 (9.68). But Breaking Bad’s S5 lives in a year where 142 dramas premiered, not 22; where the median around it is 8.16, not 7.80; and where the bottom 10% is a third of a point better than it would have been a decade earlier. The greatness is the same. The neighbourhood it lives in is different.

Look at the cloud again

Go back to the scatter at the top of this page. Now you know what you are looking at. The cloud rises a little. Its top edge does not. What grows is the cloud itself, year over year, until 2018 is so dense it practically swallows the chart.

By the time this data was assembled, scripted television had displaced theatrical film as the dominant cultural form in much of the West. Streamers were spending billions to make sure 2017’s 230 dramas became 2022’s 600. The Golden Age may be real in the data, but it was always a small effect — a quarter-point lift on top of a 30x explosion of supply.

The honest sentence about TV from 1990 to 2018 is the one critics rarely say out loud: there is much more of it, it is slightly better on average, and the very best of it is exactly as good as the very best always was.

References

Content
Assets