According to polls from the Pew Research Center, the Internet gained on Television as the public’s primary news source in 2010. Poll results are shown in their graph below.
The graph isn’t too bad, but it’s kind of busy and could use some design. Can we do better? I think so. Here’s the data as a CSV file. Get your graph on, and link to your efforts in the comments below.
You might also find the original Pew Research article helpful for background as well as more detailed data for demographic breakdowns.
One tricky point. Respondents were allowed to provide two answers if they wanted, but we don’t have the raw data, so we don’t know the exact answer breakdowns. That’s why the percentages for each year don’t add up to 100. Does this affect the design or does a small note on the bottom (like in the original) suffice?
Deadline: January 20, 2011
[Pew Research | Thanks, Elise]
Hi Nathan, thanks for inviting suggestions.
I’ve given the data a quick treatment here:
Here’s my response:
I only made subtle changes in order to emphasise the story. The original chart isn’t fundamentally bad.
Thanks for another challenge, Nathan!
I would use the same chart, but remove the data points, have thin grey grid lines, and highlight just the Newspaper and Internet lines; which have seen a decline of 31% and growth of 215% respectively – probably, a Red line for the Decline(Newspaper) and Green for Growth(Internet).
Throwing around numbers like that would get you highlighted in a Darrell Huff book 60 years ago.
One more possible thing would be to highlight the biggest increase and fall in each of the trend lines.
The biggest increase and fall across trend lines.
The Google visualization API and a motion chart is quite nice for this type of data. Here’s a demo that I set up (you’ll need to tweak the display yourself – select time for x-axis); or here’s a screengrab of the animated chart.
What about this one:
If focuses on the change since 2001, and really highlights the change in internet as a media.
Groll – thjat’s great. It really tells the story in a very very clear way. Well done. I think you nailed it.
I agree – great job on the storyline!
I added a Baseline tab to my viz so that you can see the relative change of one source against the rest a little more clearly. It’s on the same page as before:
(click on the Baseline tab when the viz loads)
Here’s what it looks like as a motion chart with Google Gadgets:
There are lots of options to play with.
Where’d the respondent numbers come from?
In this motion chart, “Respondents” are the numbers that came directly from the csv file at the top of the article. Perhaps the label should be something like “Percentage of respondents.” (The folks who collected the data allowed respondents to cite two “main” sources of news, which confused me.) I’ll make the change. Thanks!
Great idea, showing the change relative to 2001. Makes the story behind the graphic really clear.
I believe the designer can strike a balance between the information-packed graphic above a completely stripped-down version by only including the most important data in the graph, such as start/end and high/low values. Also, by using one bold color against neutrals, you can emphasize an interesting trend within the graph. Lastly, I think it’s worth examining the existing data further to come up with more meaningful takeaways, such as percent change over time.
Love everyone’s submissions! Fun exercise.
What tool are you using?
Just regular ol’ Adobe Illustrator. Thanks for the kind words!
This is a great balance. Very nice!
this is great! i’m just starting to learn illustrator — do you have any suggestions for references about how to make graphs like this, starting from raw data?
@pf – This intro tutorial could be a good place to start:
Hi pf –
I was going to recommend the one that Nathan posted above, as well as this one: http://www.smashingmagazine.com/2010/09/29/creating-graphs-with-adobe-illustrator/
The story changes in the eyes of different people. For example, for some the story could be that so many people still watch TV or read newspapers (or radio…). I think the story is that every media type is declining as main source of news, except the internet.
My 2 cents.
What I find interesting is that radio is holding its own as a source of news. This 100-year-old technology continues to be a major source of information (and entertainment), probably because it is the only media that we can safely access while commuting in our cars! It would be interesting to see whether these numbers are different for residents of New York or some other city with a comprehensie public transportation system.
I think a multivariate plot is warranted here:
Now you can see how robust radio is and how TV is the most sensitive to an increase in Internet news usage. This graph clear shows that the Internet is replacing TV and newspapers but not so much radio.
I feel like the years aren’t important enough to warrant a time series, and that presenting the data that way replaces an ugly table but adds nothing in terms of elucidation. Also, the fact that the total % changes year to year confuses a non-normalized time series.
Wow, I don’t get this one but it feels like it would be very informative if I were able to.
I like your scatter plot of scores for traditional media against the internet scores on the x-axis, and I also like the use of trend lines to illustrate the patterns that emerge – the r-squared scores show how representative the trend lines are of the data, a higher score (eg television) being better.
Taking this line of analysis one step further I’ve tried to focus on your results: http://dl.dropbox.com/u/17223586/Images/news_media.PNG
Workings, in Excel, available here:
Newspapers actually have a steeper negative slope (I reran the results myself to double check). ~-.44 vs ~.-37
That is correct. The R^2 is the squared correlation coefficient, which shows how strong the effect of Internet growth is on the decline of the others. That, coupled with the larger negative slope of newspapers, shows that newspapers lost more ground, but less of it was due to the Internet than for TV.
Another oddity is that normalizing by total responses (basically counting one person giving two answers as two people) gives this result:
The correlation for newspapers is now stronger, giving a simpler picture that says that newspapers suffered the most and that most of the effect was due to the Internet.
I’m still thinking about why the correlation is different depending on whether you normalize or not.
Just for presentation you shouldn’t label every point on each line, grid lines every 10% perhaps, maybe label first/last/high/low. As for the source data, it’s probably meaningless, regardless of the end users delivery medium, where did the story originate and how much interpretation was added as it moved from source to destination?
Here is my version that allows for a comparison of the different viewpoint options: http://goo.gl/MAvSn
Pingback: Visualize This: Where the public gets its news – Post « Another Word For It
Newspapers actually have a steeper negative slope (I reran the results myself to double check). ~-.44 vs ~.-37
Aren’t we treading on dangerous territory here by running correlations/regressions? These are aggregate numbers of data over time, not the individual responses of the entire sample right?
If you follow the link you can tell all sorts of stories. Data is cut across different demographics, and varies greatly with each cut. Something happened in 2008 that caused a huge jump across all age groups to adopt internet news. Perhaps it’d be helpful to see total subscriptions of each media across time as well. What do we know about the ease of substitution between the individual media? I could go on and on but I don’t want to derail the post too much.
Kennedy takes my vote. Minimal chartjunk, beginning and ending values, stylized formatting to place emphasis on the “outlier”, easy to read fonts. The only thing I’d add is rather than use words to label each of the media, I’d use really simple one color icons. (I’ll admit since the audience reading a Pew survey is probably different from the audience of Wired or Good, I’d probably keep the names.)
Statistics are not dangerous. It’s just math. The lack of statistical analysis is what’s dangerous. Like thinking you’re on a lucky streak in Vegas.
“Something happened in 2008 that caused a huge jump across all age groups to adopt internet news.”
These kinds of statements are exactly why kennedy’s version is dangerous. You don’t know that jump in 2008 was atypical for this time series without doing a statistical analysis. The “outliers” may not be outliers at all. Using % change may be deceptive too, since the total audience of news watchers is roughly conserved.
If your emphasis is on design and not analysis, such as with kennedy’s, I recommend being careful to present the data graphically in a simple way and not try to tell a story. Without any statistics here, your story is pretty much limited to “Internet went up and the others kind of went down.” What an anti-climactic conclusion for such a large data set.
If you wanted to break down the data by demographic, you could create more multivariate analyses, but comparing the four categories is pretty simple and has the highest statistical power since it includes all responses.
My take on the data: http://goo.gl/0O6dn
Great challenge! My makeover is here: http://www.storytellingwithdata.com/2011/01/flowingdata-challenge.html
I’m not sure about the note at the bottom. I would have loved to have seen the original data and how the end result might look if we could play with that. I like the flexibility of offering respondents two answers, but this makes what we’re looking at mean less in my book, as no accurate conclusions can be drawn from the info.
Just my opinion, but I don’t get the methodology.
I gave it two shots here. The first one is simple 2-year column. The second a stacked column.
Let me know what you think!
Rob – I really like your version. It captures the essence of the slopes on my plot nicely and can give a reader an easy way to estimate how long until the expected demise of the traditional media. If someone were inclined to include the equations for the fits, they’d be better served by presenting it as you did.
Thanks, I think, as you identified, the trend is the story, what I was trying to do was communicate the story in a sparse but arresting way: graphical, but without graphs.
I used this challenge as an opportunity to do my first charts in google docs. The visualization was more taking away the data points than anything but the other note that I would make is the scale should go from 0-100 to represent the potential range of getting news from a source.
The other way I wanted to look at the data was percent change. TV and radio fairly flat however the big story is the percent change for newspaper and internet. Its a wild shift and you can see my work in google docs:
Sorry for not adding headers and titles yet . Also if anyone knows how to get xx% to show in the y-axis instead of the decimal number it would greatly appreciated.
Throwing this out there…
Has anyone mapped major news events to the graph? From my eyes it appears that TV peaks in late Fall 2001 which obviously was a huge time for TV news in the wake of 9/11. I think Kennedy’s sort of begins to capture this (http://kennedyelliott.com/visualize-this.png). So TV appears to peak just after that at 82%.
Moving along the timeline, Internet has this nice peak in 2008 (elections?) and then drops in 09 and begins to gain again in 10.
My thesis (I used to be a TV reporter back in the day) is that news consumers have 2 modes, consumers “at rest” and active consumption (war, elections, big stories, famine, locusts..er..maybe not locusts, but certainly big storms and such.) So news consumption floats from device to device based on the news of the day.
Maybe this all seems obvious? Just thoughts. Great work everyone.
Now that’s interesting. The concept of active vs. inactive news consumption sits in the realm of a much deeper study..or, you know, a PhD. Nice thinking.
I did so in my chart: http://3.bp.blogspot.com/_T9RK3RJre_o/TTKlTUQ-EYI/AAAAAAAABus/yfH0XA23wdQ/s1600/US+news+media+by+time.jpg
My thinking was along the same lines as yours – but I couldn’t find empirical data anywhere to show which days were busier for news than other days.
I attributed the difference to the way we recall our actions – which tends to be more accurate the more specific you get. (E.g. you can more accurately tell me how many hours of TV you watched yesterday than on an average day.)
I like the “at rest” versus “active” theory too, but this data doesn’t back it up. Why do you think it boosts TV at the expense of other media? Also, if you look at the points which are very close to those of specific events, and presumably within the “active” window, they aren’t boosted for TV.
Thanks for the stimulating challenge but here’s my take: http://tinyurl.com/4f5h5qo
Guilty of splitting hairs with the data but I thought the fact that 2 responses were allowed is underplayed if only % data over time is shown, so I drew on the concept of relative media user share.
What I take away from this viz is that even if use of traditional media is declining, news of their imminent demise may be premature considering they’ve held up pretty well despite the rapid adoption of the internet. Its use may have grown tremendously but the internet has achieved parity with newspapers (though seems soon to overtake it). TV remains dominant. The 2 responses allowed is key: Seems that while a few have given up on traditional media in favor of the internet, quite a number still cling to them even as they adopt the internet. For some two-media users, perhaps traditional media is first and internet second or vice versa. It would be great to break down the data into first- and second-mentioned sources but I guess that’s for another challenge.
I love all of these great ideas! Rob Meekings, Kennedy Elliott and Synthresin’s versions really did it for me, and it was cool to see what Tableau is capable of as well in the two interactive Tableau versions.
I’m really glad I didn’t look before I did mine, otherwise I might have been too over-awed to try!
Here’s my offering – http://babblingwren.blogspot.com/2011/01/visualize-this-where-public-gets-its.html
All right, I’ll toss one in, too. Since there was a lot of great demographic data, it seemed worth including that with the top line current year, as well as the historical.
The focus on the historical isn’t so much on the actual numbers, especially since they’re hard to decipher with multiple answers per respondent, so I just summed them all up in a stacked bar graph. This makes the weight of each medium the focus.
I’ve looked a the data in absolute values and in an price-index-like way (taking 2001 as a base year).
These two ways at looking at the data hint a few interesting points.
(PS: This is the second message because it seems that the first one got spammed. Thanks)
Good work guys, really interesting to see the different takes on the information.
I’m probably a little late to the party here, but I had a go at the demographic data mentioned in the main article, since I didn’t have any marvelous revelations about the Initial question, that hadn’t already been visualized.
It’s a quick sorting of the demographic groups by news-source, and popularity of the news-sources.
That. Is a very interesting take on this. Sort of a visual overload, but you can comfortably digest everything in one place.