United States vs China: number of internet users

Here are a few charts I put together when researching the reasons for eBay's failure in China. It shows the total number of internet users in China and the United States.

As we can see, China accelerated past the United States from about 2006 onward. By 2008, China had comfortably overtaken the U.S. with 299 million users, compared with 225 million in the America.

These days, China has comfortably more than double the number of internet users than there are in the U.S. It has 571 million internet users, whereas the U.S. has 254 million users.

But the raw count of internet users only tells half the story. Or should I say 42 per cent of the story. The other factor is the penetration rate i.e. the percentage of the population that use the internet.

The chart above shows the total population of both China and the United States. The darker colours represent the number of internet users and the lighter colours represent the non-internet using population.

Only 42 per cent of people in China were internet users in 2012. However this is a staggering rise considering that less than 2 per cent of the population were internet users in 2000. The amount of headroom remaining for growth is phenomenal. No wonder so many Western companies want a piece of this big, tasty steamed pork bun.

Having been to China a few times I can personally vouch for the ubiquity of the internet and associated technologies such as smartphones. Certainly in the larger cities such as Beijing, Shanghai, Guangzhou and Shenzhen.

However, I have also experienced the Great Firewall of China at first hand. Sites such as Facebook and Twitter are inaccessible in China. Local alternatives, such as Sina Weibo, with over half a billion registered users, have cropped up as a consequence.

So, the most popular non-Chinese sites are blocked in the country. The Chinese language makes Chinese sites essentially inaccessible for outsiders. Yet, as we can see above, almost a  quarter of the world's internet users are in China. And that percentage is growing fast. Where exactly is the true internet again?


Chinese keywords: 互联网用户数, 中国, 美国
Sources: International Telecommunications Union for internet penetration figures and the World Bank for populations.

Get smart on immunisation: the stats behind the SBS article

Today, an article of mine on immunisation rates in Australia was published on SBS News. It explores the reasons why more educated parents have lower immunisation rates. The charts on the article were designed to be relatively simple to allow them to be digested by the general public.

However, for anyone that is interested in the underlying numbers and relationships, here are a couple of wonkish charts. These charts help to illustrate the relationship between education levels and vaccination rate.

First off, the scatter plot and linear regression function for education levels vs five-year vaccination rates.

The correlation coefficient between the variables is -0.28. Whilst this indicates a weak-ish relationship, I think it is one worth exploring, given that:

• The SA3-level data is the finest-grained data available for immunisation rates. These areas have populations in the range of about 10,000 to 170,000, meaning that the data is very noisy. Most areas will feature both rich and poor residents. SA4 data for immunisation would be great, but is not available publicly.
• This is social science data, therefore the data is inherently fuzzy

The chart above shows the distribution of the count of SA3s at different education levels. This chart is not population weighted. The vaccination rate for each education level is overlaid. This paints a fairly clear picture of how vaccination rates drop as education levels rise.

As an aside, I also create population-weighted charts to take into account the variation in populations seen in SA3s. The results were broadly similar, therefore in the interest of keeping the charts simple, I used the non-population weighted results.

Interactive Australian electoral division map

The map below shows Australian federal electoral divisions and the results from the 2010 election. These electoral divisions provide the basis for seats in the Australian lower house, known as the House of Representatives.

Click on the map below to navigate around the electoral seats. Click on a seat to see more details, and click again to zoom back out.

Please note: this has not been set up for viewing on the mobile version of the blog. Click here to go to the desktop version.

Loading map data...

Name of electoral division  --		Two party result -- --
Party  --    Member  --
Median total family income (weekly)  --
Median mortgage repayment (monthly)  --
Average age  --    Population  --

The results shown are based on the Mackerras pendulum, where each division is assessed on a two party preferred basis, with data taken from the Australian Electoral Commission. This assessment is based on the size of the two party margin and allows each seat to be classified as marginal, fairly safe, safe, very safe or cross bench.

This map has been built using the d3.js framework. The Commonwealth electoral division mapping data was provided by the Australian Bureau of Statistics. The political data is provided by the Australian Electoral Commission.

A heatmap of driving times in the Motor City

Here is an interactive driving time heatmap for Detroit. It visualises the time it takes to drive from Downtown Detroit to its outlying neighbourhoods. I thought it would be cool to see the artery effect of fast roads visualised on a choropleth. As the Motor City, I also thought Detroit would be an obvious choice for this.

As expected, the major roads and highways do indeed create an artery-style effect in the map below. All of Downtown is within five minutes driving time. Midtown, Corktown, Woodbridge, Eastern Market, New Center and North End are all within a mere ten minutes drive.

The map also makes it possible to spot places that are pretty inefficient to get to. For example, the Charter Township of Clinton to the north-east takes about 35 minutes to get to and is 25 miles away. This is visible as the swathe of green encompassing the township on the map. However, the Plymouth Charter Township, west of Downtown, is the same distance and takes ten minutes less time driving time.

Another interesting quirk of Detroit is that its Downtown area straddles the US-Canada border. This means that although Windsor is only on the other side of the Detroit River, driving times are disproportionate to the short distance. We can see that visualised in the map above as the darker colours emanate out to the north-west, but do so to a lesser extent across the border.

The Australian election debate: a Twitter event

The televised debate between Kevin Rudd and Tony Abbott increase the amount of Australian election tweets by a factor of about 19. The event, which started at 6:30pm on Sunday 11th August, saw over 50,000 tweets while it was in progress.

As mentioned in my previous post, I collected a lot of tweets during the Australian election debate between Kevin Rudd and Tony Abbott. Although I had expected an increase in the amount of tweets, I didn't expect such a massive increase.

The massive increase in volume of Australian election tweets is shown in the chart above. The chart shows the day of the debate, plus the two preceding days for comparison. Between 6pm and 8pm on the night of the debate, I collected a massive 71,263 Australian election tweets. The same hours on the previous night saw only 3,755 tweets sent - meaning that the debate increased the volume by about 19 times.

A key driver of this engagement was clearly the interactive elements encouraged by the media networks. Most networks, including the ABC, Channel Nine and Channel Ten had their own unique hashtags to allow their viewers to tweet their votes. This form of engagement, straddling both traditional and new media, was clearly effective for the broadcasters.

Rudd vs Abbott debate, minute by minute in Tweets

Towards the end of tonight's Australian election debate between Kevin Rudd and Tony Abbott, I captured almost 2,000 related Twitter tweets per minute, and over 50,000 were captured during the hour. This has provided me with a mountain of data - so first off, let's look at the debate minute-by-minute.

In the chart above, we can see the subjects mentioned in tweets during the debate. The first spike in tweets was about immigration, after about ten minutes of the debate. A much larger spike in activity belonged to the economic discussion, at about 18 minutes.

The economy remained one of the most popular topics of discussion for the rest of the debate, except on three occasions. In the 23rd and 48th minutes, immigration and then the environment briefly overtook the economy as the most tweeted topic.

However, with only four of the sixty minutes remaining, the subject of same-sex marriage suddenly came up. This caused a spike from mere single digits to a massive 676 tweets per minute. This was more than double the maximum number of tweets per minute seen for the economy, immigration and the environment during the debate.

The ABC and Channel Nine each asked their viewers to use hashtags to identify who they thought had won the debate. The above chart tracks the number of tweets with each hashtag per minute.

Until about the tenth minute, it was fairly even. However, it appears that the audience on Twitter started to give Kevin Rudd the edge from that point onward. Rudd's votes continued to trend upwards, peaking at 119 Twitter votes during the 56th minute - the same time as the spike in tweets about same-sex marriage.

Votes for the leader of the opposition, Tony Abbott, peaked at 47 tweets per minute on two occasions during the debate. However, generally his votes plateaued in the low-30s for the latter half of the debate, significantly below the volume of support received by Rudd.

This post was subsequently republished on the SBS News site, with the authorisation of the author.