I love lists and I hope you do too. Several lists have occurred to me recently, and I was going to ignore my itch to make them because there are probably lots of good end-of-year lists floating around now. But do any of them take a quantitative approach to list-making? I’m going to try the procedure.
How to make an End-Of-Decade List of the most Significant Days of the Decade?
Several months ago there was a widely-circulated news story about “the most boring day in history”: the finger was pointed at April 11th 1954. William Tunstall-Pedoe’s computer program ‘True Knowledge’ (an interesting phrase) identified that date, and the best I could find when skimming google for his methodology was that the decision was involved:
“being fed some 300 million facts about “people, places, business and events” that made the news. Using complex algorithms, such as how much one piece of information was linked to others…”
But is there a simpler DIY way? More importantly, if we can come up with the most boring day, we should be able to come up with the most politically significant day?
That’s the question that interests me. In the past ten years, which were the days when everyone knew that everyone was watching the news? What stories about ‘where you when x happened’ will bind a generation together?
In the next few days I fully expect to see these lists cropping up all over the place, so I thought I’d think about what quantitative political science can contribute to that process:
– What are the criteria we use to summarise ‘political events’ as such?
– How do we rank those events?
– What kinds of personal characteristics affect the way we perceive and rank ‘political events’?
I think the best way to tackle those questions is to think about them from the perspective of how they affect the different methods that I contemplated when constructing my own list:
– What about a list purely constructed from my own opinion? Hopefully it would have some value. If I was being toyed with by some evil genius and, in one of the less-sick games I was forced to play, everyone on the planet had to answer the question “what’s the most politically significant day in the past ten years?”… if my life depended on guessing the same answer as everyone else, I’m pretty sure I could do it. But where is that confidence coming from?
The next method that occurred to me was the google search. Count the hits, and rank the days that way. Simple and classic (see XKCD). But a number of issues arise with this method, especially about the validity of the measure – why should the number of google hits tell us anything about political significance? Given how I attempted to deal with the problems, it’ll be interesting to see whether the eventual product is at all sensible:
– Is there a way to design an algorithm that submits every single day to the search engine? If not, how do I decide what dates to try?
- I decided to set the agenda myself. The submission of dates then will be essentially unscientific in that it’s not easily replicable. I pick out days I think might be significant for a particular reason, and used the internet to help jog my memory.
- So how to pick the dates? This is the big question, because it’s another way of saying ‘how do I define significance’? Essentially I think I’m looking for the intersection of several sets:
I don’t promise that every included event fits those criteria, but I think any qualitative or quantitative undertaking should answer to those criteria. I submitted any date that I thought might fit that criteria: approximately 50 dates in all.
A number of subsidiary problems that were only unsatisfactorily solved:
– If I am searching for a date, do I use numerals (D/M/Y or M/D/Y) or words?
- Numerals seemed to be the best bet, since multiple languages use those numerals.
- I went for the American date formulation because I felt like that would give the biggest chance to pick up accurate readings from US websites for US events, which tend to be politically significant.
– And what time zone do I use to delineate ‘a day’?
- Many political events take longer than a day (Should the start or end date of a war be used to represent that war?) but the country where an event happened was used for the ‘date’ of that event.
– And is the number of google hits really an indication of significance? Surely it’s still biased towards ‘significance’ to people with internet access, disposable incomes, etc – just as a list composed of my own opinion would have been?
- This is not a problem that can be easily overcome, but more worrisome was the time trend. Do events in early 2001 have a chance against events in late 2010 given how much the internet has grown?
It’s this problem that is the most significant. Internet hits for 2010 events were consistently greater than those for 2000 events. I think this problem would crop up in a qualitative list too – which year do you remember better, 2010 or 2000? In this attempt at a quantitative format though, I can try to deal with it:
– Regress ‘hits’ on ‘months since Jan 2000’ and subtracting the predicted y from the actual y.
– This procedure is very rough and ready, and sensitive (as everything else is) to the inclusion or exclusion of various political/non-political events, but the adjusted formula rank does match the unadjusted formula rank in most cases.
So, here’s the top 25 political events on the list. Hopefully a decent reminder of the last decade in poilitcs:
|Date||Primary Reason||Primary Area||Category||Hits (approx)||Adjusted|
|01/12/2010||Haitian Earthquake||Haiti||Natural Disaster||813000000||564532577.6|
|08/16/2010||China overtakes Japan||China||Economic||323000000||84540062.1|
|01/01/2002||Euro Currency Introduced||EU||Economic||80700000||72412206.62|
|03/26/2000||Russian Presidential Election||Russia||Electoral||5440000||49691500.47|
|11/07/2000||US Presidential Election||USA||Electoral||20600000||44836531.38|
|10/07/2001||Afghanistan War Begins||Afghanistan||War||20400000||22119691.16|
|11/02/2002||Iraq UN Resolution 1441||USA||Conference||35600000||-208375.88|
|06/08/2004||Largest Yet G8 Summit||USA||Conference||74500000||-8843927.464|
|12/07/2009||Copenhagen Climate Summit||Denmark||Conference||235000000||-10965551.3|
|11/02/2004||US Presidential Election||USA||Electoral||80600000||-15253283.14|
|05/12/2008||Sichuan Earthquake||China||Natural Disaster||184000000||-16931870.86|
|04/20/2010||Gulf Oil Spill||USA||Natural Disaster||241000000||-17474906.98|
|11/04/2008||US Presidential Election||USA||Electoral||193000000||-22943097.67|
|03/20/2003||Iraq War Begins||Iraq||War||18800000||-27015860.42|
|03/14/2003||Human Genome Sequenced||UK||Science||18500000||-27315860.42|
|09/08/2008||LHC Switched On||France||Science||171000000||-39939355.4|
|03/14/2004||Russian Presidential Election||Russia||Electoral||26900000||-48938314.06|
My thoughts? I was surprised September 11th dropped so low, perhaps the time trend fix I attempted was even more ineffective than I feared. My further thoughts?
– 2000 was a big year for elections, and internet coverage seems to reflect what, in hindsight, seem like particularly significant elections in Russia and the US. Indeed, since every site is essentially created ‘after’ the event it’ll be interesting to see how the list changes if compiled again in another few years.
– The new pope didn’t make the top 25, neither did Hurricane Katrina, the Asian Tsunami, the Pakistan earthquake in 2008, Obama’s inauguration, the Lehmann collapse, the opening of the Beijing olympics, Benazir Bhutto’s assassination… will we say the same if the list is compiled again in the future?
– March and September seem to be big months in the Human calendar. I suppose September always seem big for those in academia, but apparently the season appeals to lots of people… terrorists planning attacks, bankers planning collapses, etc.
I might revisit this in the next few days, or write up a series of other lists I find interesting. Especially if there’s a better measure out there for specifically ‘political’ significance…
But overall, I think this is a list that gives us an interesting quantitative baseline, against which to judge lists of ‘days that changed the world’ from this decade.
 Thomas Schelling’s evil twin