March 5, 2015

Local News in a Digital Age

Facebook and Twitter—New but Limited Parts of the Local News System

As the three city surveys reveal, small segments of residents in each have added social media to the variety of ways they keep up with local news. In this section of the study, we turned from asking city residents where they get the news to closely examining social media itself. Specifically, we focused on Facebook and Twitter.

First, we looked at the Facebook pages and tweets of all the news outlets, public figures, government departments and civic groups identified in the landscape audit of each city. Second, we identified public Facebook pages and Twitter users connected to the first list.

The starting point for the Facebook study was the list of public Facebook pages collected during the audit of each city’s media landscape, such as those linked to local TV channels or municipal organizations. This list was enhanced by using Google advanced search to conduct several rounds of searches on Facebook for any public pages that used keywords referencing news in each city. That process added many more pages into the mix, from local advocacy groups to music bloggers to local journalists’ personal Facebook pages. In total, across the three cities, the center identified a total of 299 total pages, though 66 of those pages had no posts at all during the 14 days studied. (For the full list of keywords and methods, see the Methodology page).

Researchers then used the Facebook public Application Program Interface (API) to download a data file containing all of the posts from these pages during the date range for each city. This contained information about the page itself, the content of each post on the page and metadata associated with each post (such as the number of comments and likes on each post) and the text of all comments. The analysis below is based on this set of pages, posts and comments.

The Twitter analysis began with a list of Twitter handles gathered during the audit of each city’s media landscape, such as that of the local TV stations, local members of Congress and the mayor’s office. Then, using the Twitter firehose (a tool contracted through GNIP that provides full access to all content on Twitter), researchers pulled all of the tweets from each of those handles during the date range for each city. This resulted in around 4.7 million tweets across the three cities.

From there, researchers identified any account that retweeted or @ mentioned any of those handles and added this group of news sharers to the sample. This technique, sometimes referred to as “snowball sampling,” is a way to identify Twitter handles beyond structured providers of news.34 This resulted in a total of more than 4 million tweets across all three cities. Those tweets and all of their metadata are analyzed below.

Analyses of social media data are inherently messy. The second round of handles pulled from Twitter and the additional pages from Facebook were selected for their potential connection to each community, but due to the vast number of handles and pages captured it is not possible to determine how strong that connection is in every case. As such, these analyses should be treated as an exploratory, experimental addition to the more traditional survey research and content analyses contained in other sections of the report.

Facebook: An Active but Isolated Distribution Channel

Facebook Activity by CityThe multi-step analysis of public Facebook pages of news outlets, public figures, government departments and facilities, and civic groups finds that while a number of nontraditional providers compete with large legacy outlets in popularity, the stories they are covering are in many ways the same as those in other, more traditional platforms. The analysis also suggests that user comments focus on a minority of posts and tend to peter out after
the first 24 hours of a post’s life.

Most Popular Macon News Sources on FacebookIn Macon, researchers identified 65 active Facebook pages—though 17 of these did not post during the time period. From these pages, researchers collected 1,829 posts across a 14-day period from June 4 to June 18. These posts received a total of 42,744 likes, 10,825 shares, and 7,596 comments.

Most Popular Denver News Sources on FacebookIn Denver, 184 active pages were found, with 46 of those pages having zero posts within our time frame. On those pages from June 11 to June 25, researchers captured 4,579 posts, 18,330 commenters, and 37,319 comments.

Most Popular Sioux City News Sources on FacebookAnd in Sioux City, researchers identified 50 pages—three of which had no posts during the time frame. These 50 pages have been “liked” a total of 175,485 times. Between June 18 and July 2, 1,737 posts appeared on these pages, accruing a total of 18,675 likes, 5,036 shares, and 3,371 comments. A total of 1,915 users made comments on these pages.

In terms of “likes” (or followers), local TV stations and daily newspapers tend to dominate, but outliers in each city draw a following. The data also suggest that activity alone does not always lead to greater popularity.

In Macon, the local CBS affiliate, WMAZ, is by far the most popular news source on Facebook, just as in the broadcast realm. The main page has over 90,000 likes, far more than any other page—and its chief meteorologist ranks fourth with 17,767. But the page is not the most active in terms of posts: It published 234 during the time studied, less than the ABC affiliate (250) and the local daily paper (247). But these 234 posts garnered 4,547 comments, for an average of 19 per post—again far outpacing any other page.

One non-media page that stands out in Macon is that of Rep. Austin Scott. His page had 17 total posts during the time period studied, but has more likes than every news provider except CBS, and he has the fourth-highest number of comments. (Scott’s website contains no news, only press releases and thus it was not considered a news provider.) Several Facebook pages for Mercer University are also among the top 10.

In Denver, the top two pages—9NEWS KUSA and The Denver Post—together have more likes than the next 14 pages combined. While they have a similar amount of fans and posts, the Post received about twice as many comments (more than 6,700). It was the local Fox affiliate, though, that generated the most comments—11,580. The station achieved this high number by posting disproportionately about national stories. Of the top 20 most commented posts in Denver, three of the six KUSA stories were about local news. Of the 12 from Fox31, only four were about local issues.

Beyond these top sites, there are several non-legacy pages that stand out. One is the cultural weekly, Westword, which also ranks in the top outlets Denver residents turn to for arts and culture, local businesses and local community events. Westword’s Facebook page has been liked more than 78,000 times, and the 228 posts made during the two-week sampled period received nearly 2,000 comments. An even more prominent standout is the online magazine 303. Only 11 survey respondents consider it a primary source for various local news topics, but its online focus has allowed it to accumulate more likes than many of its legacy competitors: just over 59,000 likes, and about 200 comments over 80 posts. Finally, the Denver airport emerged as a surprisingly active Facebook nexus for local information. While its large number of likes (over 50,000) may be linked to the large number of out-of-town travelers who pass through, the page also posted more often—and generated more comments—than several news providers, such as Colorado Public Radio or local CW affiliate KWGN.

And in Sioux City, while the most popular page, KTIV NBC4, was the news source most often mentioned by residents as their top source for local news, the news media made up a minority of the 10 most popular Facebook pages. The remaining slots are filled by politicians such as Republican Rep. Kristi Noem (more than 23,000 likes and 738 comments over 239 posts) and civic organizations such as the school district (over 3,000 likes, but only nine comments across eight posts).

Stories Posted on Facebook Reflect Traditional Media Coverage

Across the three metro areas, the stories trending on Facebook were the same ones covered by the news media more broadly. In Macon, for instance, the firing of a deputy for a grocery store shooting had the most posts on Facebook and the second-highest number of stories in a five-day study of news coverage. In Denver, a house explosion received the third most frequent attention on Facebook, while this story was the fourth most covered in the five-day analysis. And in Sioux City, a scandal at a local casino received the second-highest number of posts on Facebook.

Audience Comments: Highly Concentrated, Short-Lived and Driven by Requests to Engage

Comments Mostly Concentrated in Few PagesAn examination of the comments across these posts reveals first and foremost that a minority of posts generate any comments: 43% do in Denver, 32% in Macon and 31% in Sioux City. Far fewer garner more than 10 comments—12% in Denver, 8% in Macon and 4% in Sioux City.

The concentration is even greater when it comes to the pages that generate those comments. As the accompanying figure shows, the majority of comments in each city were concentrated in a handful of pages. The most popular public-facing Facebook page in Denver garnered 11,580 comments, the fifth most popular page got 1,993 comments, and the tenth most popular got only 401.

Audience Engagement Is RareIn all three cities, 50% of the total number of comments came on the top three most commented pages; in Macon, 50% of the comments were from the most popular page alone (13WMAZ.com).

For those posts that did attract at least one comment, more than 85% were left in the first day after posting. Additionally, less than 15% of commenters left more than two comments in the week we studied. If individuals are having conversations on Facebook, it does not appear to be a frequent occurrence in these official forums.

While there is no clear pattern to the kind of post that generates comments, one frequent element was that local news generated more interest than stories from outside the area. Of the 100 most commented posts in each city, between half and three-quarters (58% in Denver, 71% in Macon, 54% in Sioux City) are about local news.

And in Denver, 50% were spurred by some form of audience outreach, such as questions, requests for photos or videos, and polls. But this did not seem to be much of a factor in the other cities: This was true of just 28% of the most commented posts in Sioux City and 20% in Macon.

The types of engagement put forward in these posts varied. The most commented post in Denver, from the local CBS affiliate, asked “Would you like to see Hillary as president?” (564 likes, 639 comments), while the third most commented post asked if residents would be interested in trying a very large waterslide (1,538 likes, 615 comments).

Most Pages Allow Public to Post, but Few Take Them Up on ItAnother, arguably higher, level of interaction is to allow users to post their own stories on the Facebook wall of a provider. A resident, for example, would be allowed to post about a news issue on The Denver Post’s Facebook wall—a post that all of the 300,000-plus people who follow the Post on Facebook could potentially see in their newsfeeds. While most pages allowed members of the public to make posts (89% in Denver, 68% in Macon, 76% in Sioux City), the majority of posts were from the owners of the page, not members of the public (77% in Denver, 88% in Macon, 73% in Sioux City). Whether this is due to few submissions, site moderators that do not approve them or some combination of the two is unclear—though the end result is few pages with much citizen content.

Most Comments Made in First Few HoursHowever, the exceptions showcased the possibility of Facebook as a public information hub—or at least a page filled with more external than internal content. The page of South Dakota Rep. Kristi Noem was filled during the time studied with 230 posts from citizens, compared with just nine from the page’s owners (though Sioux City is in Iowa, the Sioux City metropolitan area includes parts of South Dakota and Nebraska as well). And the conversation that emerged through those posts was about politics more generally rather than about Noem’s own actions or policies. For example, citizen posts included a link to a YouTube video on immigration; a photo of the results from a Quinnipiac poll; quotes from the Hobby Lobby court decision (without a link); questions for Noem about the war in Syria; and a link to a Yahoo story about Congress’ work schedule.

Mostly Self-LinksAt the same time, however, the majority of these user-generated posts came from a handful of people. While 14 users made posts on Noem’s page over these 14 days, 80% were made by just three people, suggesting that the number of individuals who take advantage of these kinds of options can be quite limited.

One other distinction on Congresswoman Noem’s page is the degree to which even her own posts linked to outside content. Of the eight links posted by Noem in the time period, just two went to her own site. Contrast this to her fellow member of Congress, Steve King, whose official Facebook page features no citizen posts. Of the nine posts on his page with links, six went to his own home domain.

The distinctions of this one page had a large impact on the breakdown of overall Facebook content coming from Sioux City news providers. Across the 50 pages studied, fully 69% of links went to outside content. That is four times the percent of links from Macon news providers and also more than Denver. Instead, more than half of links from posts made on the pages of Denver news providers went to their own websites; in Macon, the figure was even higher (83%).

Most Posts Contain More than Text

A Facebook post can have two different elements to it: a text-based post (such as posting “Happy birthday!” to your friend’s wall), and non-text content, such as a link to a news story or YouTube video, or embedded photo or video (made by clicking “Add photos/video” while composing the post). A link or photo can be posted by itself, or it can be posted with accompanying text that might explain, provide context or add information.

Type of Content SharedIn general, posts rarely contain only text (18% in Denver, 30% in Sioux City and Macon). But the cities differ in whether this additional content stands alone or is paired with some explanatory text. In Denver, the most digitally adept city in our study, about three-quarters of posts use both text and additional content. In Sioux City, however (which was generally less engaged with the digital space), about 20% of posts contain only this additional content, such as a photo or link without any added text by the poster.

This additional content consists primarily of photos or videos in Denver and Sioux City, but primarily of links in Macon. Overall, six-in-ten posts with additional content in Sioux City and five-in-ten posts in Denver contained photos or videos. In Macon, however, only 41% of posts with additional content contained photos or videos, while 56% contained a link.

In two cities, the post with the most likes is a photo with accompanying text. In Denver, it is a post from the Denver Police Department with a picture of a baby who’d been rescued and cared for by officers. In Sioux City, it’s a picture of a double rainbow posted by local CW affiliate KTIV; given the recent flooding, the picture had special resonance for residents. In Macon, the most liked post is text only: an update from WMAZ anchor Frank Malloy on the health of his wife, who was in the hospital after a car accident. (She was later charged with DUI in connection with the accident.)

The Twitter-verse in Denver, Macon, and Sioux City

The more public nature of Twitter compared with Facebook allows for a different kind of analysis focused more on the organic ways in which local news providers and residents use the platform.

The Pew Research Center, through a contract with GNIP, has access to all Twitter content—known as the Twitter “firehose.” One of the biggest challenges in studying Twitter is narrowing in on posts stemming from each geographic area.

Researchers first identified the Twitter handles of all news providers identified in the landscape audit of each city. These included news organizations, individual journalists, government entities, schools, neighborhood associations and any other organization that produces original local news content. Then researchers pulled from GNIP all the tweets generated from those handles over a five-day period for each city.35

Tweets in the SampleFrom that corpus of tweets, researchers made a list of all new handles that @ mentioned or retweeted any tweet from the first round data and then pulled all of the tweets from this second, and entirely new, set of handles and added them into the first data set. That amounted to a total of nearly 5 million tweets and over 30,000 Twitter users. There were 409,868 total tweets in the Macon sample, 1,310,843 in Sioux City, and 3,201,804 in Denver.

The goal of this process was to collect a group of Twitter users (and their tweets) who are connected to each community. To define “connected,” we collected Twitter users who interacted with our original list of news organizations, civic leaders and other information providers in each city.

Twitter Users in the SampleFrom here researchers worked to refine the list to get as close as possible to what could be identified as local content. This involved analyzing the URLs contained in tweets to better understand the stories being shared as well as looking at what hashtags were being used. In Macon it was possible to review the tweets individually as well, but this kind of analysis was impossible in Sioux City and Denver given the volume of tweets.

Finally, to facilitate the analysis, researchers separated posts that contained URL links and those that did not. Dividing the tweets this way served two purposes. The first was to understand how news stories are shared on Twitter. By looking only at tweets that included a URL, researchers were able to better understand how local news outlets are shared on Twitter and how national outlets compare. The second was to simplify the process by isolating the more difficult to analyze tweets that did not include a URL link. Researchers looked at hashtags that were used and read the tweets that were shared the most, though it was clearly not possible to read all the tweets included in the massive data set.

Overall, the analysis found little discussion of the local news stories covered most by the local news providers. Instead, conversations tended to focus on content that would not be classified as news—such as conversations between local residents. When posts did relate to news and information, they were more often national in scope than local—and most often tended to be political in nature. The top national URLs are all the same stories that were in the national press that week, and the top hashtags in tweets without URLs were all nationally oriented hashtags: #tcot, #bringbackourmarine, #renewui, #HR803, #immigrationreform.

One standout story occurred in Macon, where a local pop band participated in a VH1 contest.

In Macon, One Major Local Thread Emerges

For the week of June 6 through June 13, 2014, the Twitter conversation around Macon focused heavily on a single story. A local band called Good Night Alive was competing in a VH1 contest called “Make a Band Famous.” The show aired on Wednesday, June 11, in the middle of the week studied in Macon, and part of the band’s success in the contest depended on how many users tweeted about the band.

Of the over 400,000 tweets studied, researchers identified 77,427, or 19%, that were about Good Night Alive.36 No other subject came close to that number of tweets.

Of the 77,427 tweets about the band, 12% contained URLs. The most shared was a link to a song by the band hosted on Soundcloud (shared 3,486 times), and the second was to the band’s section of the VH1 contest page (shared 2,302 times). These were also the most shared URLs for the entire week studied—and shared far more than any other story in Macon.

No other story, local or national, came close in size to the story about Good Night Alive.

The next most shared URL after the band was a story by the National Council of Resistance of Iran about the Iranian leadership. It was shared 1,957 times.

Third, at 1,102 shares, was not a news story at all but a link to happybabyworkout.com, a site about helping mothers with young children exercise with their kids. After this came a link to a tweet voicing opposition to a bill in Congress that would require the Department of Agriculture to not issue licenses to employees at horse shows, sales or auctions that “sore” horses.37

As noted above, it was challenging to find any local stories, other than the stories about Good Night Alive, during the week studied. The next most shared local story was shared just 22 times during the five-day capture. The story, posted by a local TV station, was about a full moon visible on the Friday of our capture dates and was shared 24 times during the capture. All of the other local stories that were shared using a URL appeared only one time during the time period studied. These included a story about a local party in downtown Macon on June 28 and a link to the site middlegeorgiaceo.com that featured a story about challenges facing businesses in the Macon area.

Hashtags Point to a Similar Focus as URLs

While the main focus of this analysis involved the study of tweets containing a link to a website, nearly half of the tweets that were collected in Macon (44%) did not include a link. This side of the Twitter conversation is much harder to study. Twitter users in the three cities fell into the sample because they at some point retweeted or engaged with one of the news or civic accounts identified, but because all their tweets were pulled in, many conversations that were not about the news were also included. And many of the tweets were parts of longer conversations and thus it was hard to decipher their meaning.

Despite these challenges, there are some telling statistics about this group of tweets. While they did not include a URL, many of them did include hashtags. Here again in Macon the story is clear; the Good Night Alive story was also dominant in tweets without URLs. In total the combination of #mabf and #goodnight live was used more than 65,000 times (this includes various spellings and capitalizations of the two hashtags).

The other popular hashtags in Macon were nationally relevant hashtags such as #renewui, which means “renew unemployment insurance,” or #bringbackourmarine, a hashtag campaign to bring back a U.S. Marine who was being held in a Mexican prison.

Sioux City: National News Supersedes Local

In many ways Twitter activity in Sioux City over the week studied resembled Macon’s—aside from the Good Night Alive story thread. Among the tweets captured, the top shared news is national, local news is shared rarely, and Twitter is used as often for conversation between users as it is to share news.

The local content with the most shares on Twitter was weather, with a link to a live weather cam from Iowa Weather Now (that link appeared 24 times). The next most shared stories in Sioux City were three stories that each appeared eight times over the week studied: a story about a Sioux City resident who won a gold medal at the Special Olympic Games, the Hard Rock Casino assembling a giant guitar, and an obituary for a local sports icon.

Compared with national news stories and other content, these local stories appeared far less often. The top shared URL during the time period was a Rock the Vote effort by votelatino.com that was shared 1,600 times (the site has since been taken down). The top news story URL was an article from The Hill about how Republicans in the House killed the latest immigration reform bill. In fact, of the top 10 URLs by number of times they appeared during the week studied, all either linked to a story about a political topic or linked to a Twitter status about a political topic.

Hashtags Cross City Boundaries

Of all the tweets analyzed in Sioux City, 42% did not contain a URL. As mentioned above, this lack of a URL makes it harder to judge what the full set of tweets is discussing (it totaled around 550,000 tweets).

Of tweets that did not contain a URL, 44% were original posts and 56% were shares. This rough pattern fits with Macon, as does the number of @ mentions. More than 90% of the tweets without URLs contained an @ mention.

In Sioux City the top hashtag was #renewui, a hashtag that was popular in Macon as well. This refers to the effort to renew unemployment insurance. Other top hashtags in Sioux City were related to this topic such as #hope4jobs and #HR803 (referring to the bill in Congress that would renew unemployment insurance). All of these hashtags grouped together were tweeted more than 100,000 times—9% of all the hashtags used during the week studied.

Like the top URLs, there is a very long tail of hashtags. Of the 84,404 unique hashtags that week, 51,234 (61%) were used a single time.

Other top hashtags included #TCOT, which is a common hashtag on Twitter used to denote conservative political tweets. As in Macon, #bringbackourmarine was a common hashtag referring to a U.S. Marine being held in a Mexican prison. Finally, hashtags referring to immigration reform appeared many times in the most used hashtags. There were no hashtags shared more than once that referred directly to local events in Sioux City.

Denver: Local News has a Somewhat Larger Place in the Twitter Talk

In keeping with Denver’s more digital nature, local news appears more often there and is more likely to be shared on Twitter than in the other two cities studied. That said, nonlocal news and other content still outpace local news in terms of sharing, even in Denver.

A local sports story about a player for the Denver Broncos who was incorrectly reported as being arrested in Dallas led the list of shares in Denver. That URL appeared 237 times.

The next local story—about a shooting at an outdoor concert venue outside Denver called Red Rocks—also appeared on mainstream local outlets like local TV. The story was covered in 11 local outlets (and their websites) and was shared 116 times on Twitter. Most of the conversation about the shooting at Red Rocks included a link to a single Denver Post story.

Still, national and other nonlocal news far outpaced local news among this group of Twitter users connected to the Denver community. The top URLs during the week studied were mostly non-news URLs, with the exception of a Washington Post article about various Republicans talking about the possibility of impeaching President Obama. The top shared link overall was a Rock the Vote effort at a site that targets Latino voters called votelatino.org (the same URL as in Sioux City).

This pattern shows that while local news rises to the top more often in Denver, it is still competing with many other content types on Twitter. As in the other two cities, political topics rose to the top of stories that were shared the most on Twitter in Denver. Unlike the other two cities, immigration reform was the center of the national political topics being discussed in Denver. The URL that appeared the second most often was to a site that allows users to easily find out who their representatives and senators are in Congress and contact them directly about immigration reform.

In terms of hashtags, the 50 most used were dominated by national politics. However, a few were explicitly local. #Denver was the 24th most used hashtag overall, with 6,432 instances. #copolitics came in 38th with 5,159, and #cowx, a hashtag for discussion Colorado weather, was 45th with 4,584.

Complicating matters in Denver was the fact that two national issues were of particular local interest: marijuana legalization and immigration. The hashtag #marijuana was the 40th most used hashtag in Denver, with 5,018 instances, though we cannot say for sure what percentage of these tweets were strictly political. And of the top 50 most used hashtags, nine were about immigration, by far the biggest single subject addressed in these hashtags.

Tweets Without URLs in Denver

In many ways tweets in Denver without a URL look very similar to those in Macon and Sioux City. In tweets without URLs, 28% were @ replies. This seems to be the most common alternative to including a URL in a tweet. In tweets that did have URLs, only 2% were @ mentions.

As with Macon and Sioux City, many tweets without URLs discussed local issues using hashtags. The 50 most used hashtags in tweets without URLs included only one of the three explicitly local hashtags (#copolitics, 1,662 tweets, #42), and the semi-local #marijuana did not appear.

As in Macon and Sioux City, the nationally relevant subject of immigration, however, was more prevalent: 12 of the 50 most used hashtags in tweets without URLs were about immigration.

  1. Twitter does provide some geotagging for tweets and location information for Twitter users. Researchers analyzed a random sample of tweets and attempted to use keywords to identify Twitter users in each city. However, the vast majority of self-identified location information did not identify where the user was located in any meaningful way. Another potential location method is through geotags that Twitter allows users to voluntarily add to their tweets. This setting is turned off by default, but users can opt-in and have their tweets automatically geotagged as they are posted. This option, however, is used by a very small percentage of Twitter users. General estimates are around 2% of Twitter users—and researchers found that in these three cities that percentage was even lower, making any resulting mix of handles extremely limited. There was, then, no optimal way to use geographic information to gather a group of Twitter users in each city.
  2. Denver was studied June 16-20, 2014; Macon, June 9-13, 2014; Sioux City, June 23-27, 2014.
  3. Researchers filtered the text of the tweets, the URL, and the hashtags for the terms “goodnightalive,” “VH1,” or “MABF,” which stands for “Make a Band Famous,” the title of the VH1 show featuring Good Night Alive.
  4. The language of the bill according to the Congressional Research Service summary is: Prevent All Soring Tactics Act of 2013 or the PAST Act— Amends the Horse Protection Act (HPA) to direct the Secretary of Agriculture to prescribe regulatory requirements for the Department of Agriculture (USDA) to license, train, assign, and oversee persons who are to be hired by the management of horse shows, exhibitions, sales, or auctions and are qualified to detect and diagnose sore horses or otherwise inspect horses at such events. (The soring of horses refers to the application of blistering agents, burns, lacerations, sharp objects, or other substances or devices to a horse’s limb to produce a higher gait by making it painful for the horse to step down.)