Posted by on Apr 25, 2013 in Research & Data
つぶやいてください: Twitter’s Networks in Japan

This past summer, Japan witnessed strong political response to the local energy crisis. In March 2011, the Tohoku Earthquake devastated the nation, and since then a primary topic of conversation has been whether Japan should reinstate its nuclear energy program. As many well know, Twitter became a crucial communication technology used widely during the crisis.

[flickr video=5884626815 secret=d851598b9a w=640 h=360]Twitter messages and retweets emerging after the 2011 Tohoku earthquake

While Japan has been known for its own local social media platforms (Nico Nico Douga for video sharing, Mixi for social networking and blogging, and 2channel for subculture), in the past few years Twitter has grown immensely in popularity and adoption within the country, with thousands of Japanese “mumbling” (つぶやき) daily. From books chronicling the platform’s spread to television dramas about using the service, Twitter has become a social media phenomenon within the country.

Unfortunately, we don’t know too much about how Japanese individuals, groups, and companies use Twitter. For one, we know that these users are very active: Japan has set records during New Year’s celebrations for volume of tweets per second produced. Conversely, last year, Twitter’s engineering department released a report that showed that – compared to other major cities around the world – Tokyo tweets just as much during the night but less so during the day.


A report by Twitter’s engineering department suggests that Tokyo tweets much less at night than other cities.

To see how true the “Tokyo only tweets at night” hypothesis was, we first looked at language use on Twitter over the period of one week. As you can see below, most languages – except Japanese – exhibit diurnal peaks (meaning one peak in the morning and one peak in the evening). Instead, Japanese has a much less pronounced bump in the morning, sloping up to high activity in the evening. It seems likely that social norms, especially around using the internet at work, may shape these behaviors.


One week of languages on Twitter.

We then randomly sampled users from both New York City and Tokyo to compare them. Below, you can see graphs for a few days of these users’ volume of tweets. Though we would expect to see diurnal peaks, we see that both cities exhibit more tweeting at night. Still, it remains clear that Japanese users tend to use Twitter more at night, and these behaviors of course greatly impact how information can travel across the various networks that make up Japanese Twitter.


Three days of activity from 5000 users each from Tokyo and New York City.

When looking at possible differences in other data in the comparison between Japanese and American Twitter users, we did not find much dissimilarity. For example, looking at followers, American users had a slightly higher frequency of accounts with 0 or 1 followers compared to Japanese users, but both samples showed long-tail trends in their follower counts, ending in the hundreds of thousands of followers.

One distinction we did find was in how users tweet. While it’s to be expected given phone culture in Japan, Japanese users in Tokyo also exhibit a higher percentage of mobile use compared to their New York cohort. Web use is much higher in New York City compared to Tokyo, and Japanese users also illustrate a much more diverse set of third-party platforms.

Percentage of platform use for users in Tokyo and New York City.

Generally, we also saw that many Japanese businesses are unlikely to use Twitter to reach their audiences. A study by Adam Acar shows, for example, that uptake of Twitter by top Japanese companies has only reached 60%, whereas 95% of the top 100 American companies are active on the platform. Also, while 86% of American brands tweeted during the study’s analysis, only 41% of Japanese brands did the same. Still, one large Japanese retailer we examined, Rakuten (@RakutenJP), has shown that occasional Twitter campaigns do lead to greatly – and quickly – increased follower counts.

url

Rakuten follower counts starting summer 2012 show dramatic increases in followers after campaigns.

Beyond its popular use, Twitter has now established itself as a public sphere, a forum for communication where ordinary citizens (as well as journalists, politicians, and celebrities) can discuss and share information about critical civic topics. Over the summer, we decided to look at how Japanese Twitter users were speaking about and spreading awareness of these crucial issues: in particular, we investigated the conversations and social networks around the antinuclear protests across Japan. From July 10 to July 28, 2012, we looked at tweets containing the word “protest” (抗議) to see what kinds of discussions, markers, and communities would turn up. We knew that each Friday a scheduled anti-nuclear protest was held, and we wanted to see if we could map individuals who chatted about them.


Network graph of topics that emerged from the protest dataset.

The colors in the nework graph represent the topical clusters that emerge from the data. The largest cluster (9.52% of the nodes) revolves around popular journalist Yasumi Iwakami (岩上安身) and his journalistic enterprise, the Independent Web Journal (http://iwj.co.jp/). Iwakami-san was a popular voice in the criticism around the antinuclear and radiation issues in the country. The second-largest cluster (8.94%) revolved around the activist hashtag #紫陽花革命 (“hydrangea revolution”) and involved other key hashtags related to the antinuclear protest, such as #脱原発 (“abandoning nuclear power generation”) and #再稼働反対 (“reoperation opposition,” referring to the reopening of shut-down nuclear plants).

Other related clusters emerged: many people mentioned NHK’s coverage of the protests (5.48%), a large community formed around a general discussion of the protest demonstrations (4.64%; #原発, “nuclear power plant”; #デモ, “demonstration”; #genpatsu, a romanized version of 原発; etc.), and in particular the massive protest in front of the Prime Minister’s residence (4.62% with #首相官邸前抗議, “protest in front of prime minister’s official residence”; 3.7% with #官邸前, “in front of official residence”; 4.02% with “Ustream,” which livestreamed the residence protest) . One meme also appeared (in the 4.02% just mentioned), where @monjukun, a mascot that became well-known for tweeting in easy-to-understand terms against the government’s nuclear policies, became an important individual in the debates.


Close-up of core topical areas from the protest dataset.

However, since the keyword 抗議 (“protest”) is fairly common across any political movement in Japan, we also found other emergent events and conversations. For instance, the Trans-Pacific Partnership Agreement, a copyright agreement between Japan and the US known colloquially as TPP, occupied 5.36% of the discussion. Even more, a cluster emerged around the hashtag #抗議 (“protest”), which was used for a variety of activist purposes on Twitter. In particular, the hashtag was commonly paired with “FAX,” relating to a fair number of calls to spread word about particular protests via phone and fax. “FAX” was especially pertinent in the middle of the dataset, as many Twitter users spoke about protests regarding the suicide of a schoolboy in the city of 大津 (Otsu). Finally, another general topic emerged around #政治 (“politics”), #seiji (the romanized version of 政治), #マスコミ (“mass communication” or “the press”), and #民主党 (the Japanese Democratic Party).

Since we looked at tweets over a long time period, the graphs below illustrate how various topics emerged over the course of the 18 days. The weekly spikes – particularly around the power plant hashtags – match up with the weekly protests held every Friday. What is perhaps most surprising about these spikes is the lack of a conversational spike around July 16th, when tens of thousands of Japanese protestors, in a special day of protest, congregated in Yoyogi Park in Central Tokyo, shown below. The lack of social media coverage of this event suggests that the “tweeting at night” behavior common to Japanese users may have actually affected Twitter’s potential to bring attention to this gathering.


Antinuclear protest in Yoyogi Park.

Regarding the social network underlying these conversations, the dataset covers a broad range of users. It is important to note that most of the users who participated in protest conversations were Japanese-speaking users, and the social networks reflect Japanese users connecting to other Japanese. The interesting aspect of the protest networks is that, generally, the network covers a wide range of participants, generating discussion amongst the core groups involved (ie., journalists, academics, and news critics [right side of graph below]) but also spanning to other strong Japanese Twitter networks, like very active netizens and otaku (media fans) [left side of graph below].


Sample of 10,000 users from “protest” social network.

However, if we look at the network of users that participated in antinuclear protest discussions as a subset of the graph above, these networks were very dense. In other words, the network of users talking about antinuclear issues acted as a sort of echo chamber, where tweet volume might have been high around the topic but where these discussions didn’t spread to other Twitter networks in Japan.


Users who mentioned #genpatsu or #原発 hashtags in their tweets.

Above, you can see that users who tweeted about the nuclear power plant (genpatsu) form a very dense network. While the discussion was vibrant amongst these users, the potential for information to move across networks and spread widely throughout Twitter was not strong. The image literally illustrates a bubble of conversation.

Overall, Japanese Twitter user continues to grow, so we will see many interesting opportunities for Twitter to act as a platform for discussion and marketing in the coming months. Anyone considering to use Twitter for a campaign should consider that particular discussions might become caught in a dense network, so they should strategize about expanding across a diverse set of users. Also, information may travel better at night, when users are more active. Finally, mobile still remains strong in Japan, though the ever-increasing use of smart phones (instead of traditional Japanese “keitai” phones) may challenge companies to design projects that bridge these users.

Comments

comments