3 February 2009
You can now answer that question and so many more. The Japanese Statistics Bureau conducts a survey every five years called the "Survey on Time Use and Leisure Activities" where they give people journals to record their activities throughout the day. Thus, they have a survey of what people are in Japan at any given time of the day. This is fun data in of itself, but it was made downright addictive by Jonathan Soma who created a slick Stream Graph based on the data. (via kottke)
There are actually three Stream Graphs: one for the various activities, another for how the current activity differs between sexes and a final for how the current activity breaks down by economic status. Thus, the view contains not only information about daily routines, but also how those routines vary across sex and activity. For instance, gardening tends to happen in the afternoon and evening at around equal intensity and is fairly evenly distributed between men and women. Household upkeep, on the other hand, is done mostly by women and mostly in the morning. This visualization is so compelling, I think, because it allows for deep exploration of rich and interesting data (to be honest, though, I find the economic status categories a little strange and not incredibly useful).
I think there are two points that come to mind when seeing this. First is that it would fascinating to see how these would look across countries, even if it was just one other country. The category of this survey on the website for the Japanese Bureau of Statistics is "culture." Seeing the charts actually makes me wonder how different this culture is from other countries. Soma does point out, though, that Japanese men are rather interested in "productive sports" which is perhaps unique to the island.
Second, I think that Stream Graphs might be useful for other time-based data types. Long term survey projects, such as the General Social Survey, track respondent spending priorities. It seems straightforward to use a Stream Graph to capture how priorities shift over time. Other implemented Stream Graphs are the NYT box-office returns data and Lee Byron's last.fm playlist data. This graph type seems best suited for showing how different categories change over time and how rapidly they grow and how quickly they shrink. They also seem to require some knowledge of Processing. There are still some open questions here: What other types of social science data might these charts be useful for? How or should we incorporate uncertainty? (Soma warns that the Japan data is rather slim on the number of respondents)
Posted by Matt Blackwell at February 3, 2009 5:37 PM