Valar Morghulis

Introduction:
Game of Thrones, for those of you who are not familiar (I am really not; see disclaimer), is famous for high mortality rates that make viewers and readers very angry. As a global health graduate student (with a lot of free time at the moment…) I became interested in figuring out precisely how bloody this universe was, whether deaths varied by gender and status/occupation/affiliation, and how mortality in Westeros compares to mortality in low- and middle-income countries. The professor for a course I’m TAing in a few weeks also mentioned that he wanted an assignment on life tables, so, you know, I figured I should learn what they are.

No character names are used in this blog post, but I guess there might be spoilers if you can back-calculate in your head.

Methodology:
All age, gender, and affiliation data was obtained from a combination of the Game of Thrones Wiki and A Wiki of Ice and Fire, based primarily upon the shows. For the purpose of this analysis, characters who are missing in the films but alive later in the books are listed as alive, characters who are missing but presumed alive for some reason are listed as alive, and all other characters of unknown status are listed as dead.

Characters of unknown age have been treated in one of two ways:

1) Characters with an unstated age but whose births and deaths can be referenced back to other events are tallied using those references, typically marriages, births, and battles;
2) Characters for whom a reasonable estimate is available have been rounded to the nearest 5 year interval;
3) Characters for whom no estimate is stated have been replaced with the age of the actor or the actress at time of filming;
4) Characters for whom no age is known and for which no estimate is available have been excluded from the analysis (see raw data);
5) Characters who have come back to life multiple times and/or transformed into some nonhuman sort of being are ignored because, come on. Dragons, direwolves, etc. are also excluded.

For the purposes of calculating survival rates, I excluded characters younger than the age in question who are not yet dead so as not to discount their potential survival. For example, I’ve logged 47 affiliates of House Stark in total. 39 are 15 years of age or older (as of year 300). Five Starks died prior to age 15. To obtain the proportion of Starks who survived to 15, I want to ignore the other 10, so the calculation would be 39/(39+5) = 88.6% survival.

Most of this was done in Excel and STATA.

Results & Analysis
Suffice to say, (almost) nobody grows old in Westeros. Here’s the big one:

main chart

In short, fewer than 50% of characters in Game of Thrones live to see a 45th birthday. Only 12.9% would ever become eligible for Social Security in the United States.

Breaking it down:

Gender differences actually aren’t too…stark.

GoT Gender

Women fall off sharply in two age groups: late teens/early 20s (marriageability/maternal mortality/choosing loyalties via husband), and early 40s (Expiration? Age of onset for family-wide assassinations? Who knows).

I found this surprising – though there aren’t terribly many female characters (I tallied 92), I expected men to fall off in droves at the battlefield while their wives stayed home alive. It’s possible that this is still the case and that the homebodies aren’t intriguing enough to discuss (or that extreme gender bias in Westeros leads to sex-selective abortions).

Differences between Houses aren’t particularly interesting, either. For a sample, have a look at the Starks, Baratheons, and Targaryens:

House SBJ

Different drops during the battle-heavy years, but no clear lead. The Night’s Watch makes for a more interesting comparison. Let’s take a look at Jon Snow’s decision:

Snow Dilemma

Obviously they have to make it to fighting age to join, so Watchmen miss out on early childhood mortality. But they quickly catch up, outpacing the normal Starks in deaths during childbearing years, catching up as those in the House head to battle and the better of the Watchmen survive, and then stripping away once more toward slightly earlier deaths. This difference, also, was not as well-defined as I had anticipated.

Modern-day comparisons:
Despite a promising early start, the battle- and intrigue-heavy adult years in Game of Thrones lead to a relatively linear decline in survival unlike anything we really see on the country level. Despite this, a sprightly few nonagenarians sneak their way into the plot at a rate that outpaces most low- and middle-income states today.Comparison Table

In graphical form:

Comparison Graph

“Hold up,” you say. “You’re meaning to tell me that the infant mortality rate in Game of Thrones is lower than Afghanistan’s?” Well, I guess, but George R.R. Martin is not about to name every two-year-old struck with pneumonia at King’s Landing. Which brings us to…

Conclusions:

All of this data is obviously very limited by the nature of narrative. Characters who are interesting get named, discussed in text, and logged on Wikipedia. They are usually not babies, and in fantasy, from what I’ve gathered, they are frequently wise old men/women/weird crow-men/whatever whose more boring contemporaries have all died off. Authors don’t name everyone who dies on the battlefield because then we would never get past it.

That aside, there are a few interesting take-aways.

First of all, despite battlefield tasks expected of men, it is not good to be a woman in Westeros. You don’t necessarily have to suit up for war but but you aren’t particularly protected by those who do, you suffer all the standard childbirth-related maladies, and yet you still don’t get to enjoy gender-based advantages in longevity seen nearly everywhere else. Cheap deal.

Second, this is probably not a good method of predicting anything about future plotlines whatsoever. House-based fatalities come in sweeps.

Third, I don’t think this mortality pattern is reflected anywhere in this modern world at a level anyone actively monitors. The nearly-linear decline in population with age isn’t characteristic of even developing countries, where you’ll typically do pretty alright until your 60s if you make it to adulthood.

Finally, I definitely do not want to live there.

—-

Disclaimer: Everything I know about the Game of Thrones universe came from writing this post and the ~5 episodes I drunk-watched in college. There are absolutely classification mistakes; I’ve never read the books or watched the series, so please don’t yell at me. Feel more than free to make your own improvements (though I’d appreciate credit if you use my derived materials)! Here’s a link to the raw data I compiled. Let me know if you spot obvious errors in methodology and I’ll try to update this appropriately.

Let me know if you improve upon this or do your own and I’ll link it here! Brett Keller of Hunger Game Survival Analysis Fame is already working on one, and that will be much more rigorous.

Valar Morghulis

8 thoughts on “Valar Morghulis

  1. As I commented on Twitter, this gets censoring correct, which is good, but there’s still an issue with left-truncation. As you note for the infant mortality issue, people who die at two aren’t going to become noteworthy later.

    Ideally, you could get an entry time for each person, the first age at which they would have been mentioned regardless of their age of death. The lifetable approach handles this easily — people appear in the denominator from their entry time until their death. The problem is that it’s not feasible to work out the entry times — it would require reading GRRM’s mind.

    What might be feasible and worthwhile would be to estimate infant mortality from, say, first-born children of named couples. Or to just assume it’s what you’d expect from the sanitation and public health systems.

    1. It occurs to me that there is a partial solution. One could use the first age at which a character was mentioned as the entry time. There would be a loss of information, but the bias would be reduced.

      The bias wouldn’t be eliminated, because it’s possible that GRRM decides on page 19303 of a book to add some more corpses and then goes back and inserts them earlier in the story.

  2. There is nothing to yell at you about! I’ve read all the books, seen the show, and I find your analysis informative (and I’ll admit, entertaining.)

    I’m particularly happy to see your analysis of mortality between the main houses of Stark, Baratheon, and Targaryen, but I’d have wished House Lannister was in there too. (Not yelling, just observing.)

    Thanks for sharing this, I’m always happy to see Game of Thrones data used for academic purposes.

  3. […] Jordan Sheremerhorn, a Global Health Graduate from Duke University looked into the wikis for HBO’s Game of Thrones and for the novels, A Song of Ice and Fire (ASOIAF) and she found that there are very few elderly people, with less than half the characters reaching the age of 45…Similarly survival rates vary greatly between a character’s house affiliation (just think about the number of dead Starks there are). […]

  4. Devon Allaby says:

    This is great. Would you mind posting what affiliation each number in the raw data represents? I can guess at most of them..

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s