Ep.52: When Your Boss is an Algorithm
With Emily Guendelsberger and Sean O’Brady
Building on The Shattered Mirror report of January 2017, the Public Policy Forum continues to study the news environment in Canada with a mind toward understanding better, and offering solutions for, how to sustain journalism and how to challenge the growth of misinformation in the new digital public sphere. We view these as separate but related strands, both in need of attention. This report sets out recent research findings about widening gaps in local and community news. A companion report, What the Saskatchewan Roughriders Can Teach Canadian Journalism, discusses emerging ownership models.
Our empirical research, in summary, has found that the number of newspaper articles appearing in a sample of communities in all five regions of Canada fell by almost half over the last decade. Coverage of democratic institutions and civic affairs declined by more than a third. This drop coincides with the downsizing, mergers and closures of print news outlets. We saw a similar pattern across all 20 communities studied: small and medium-sized; anglophone and francophone; and, significantly, those with and without newspaper closures (although naturally more pronounced in the latter). There were fewer articles per year in every community examined.
Coupled with previous PPF research and the work of academics and research institutions in Canada and elsewhere, these trends support the conclusion that the dominant for-profit models of recent decades are not sustainable and that the quality of informed democratic discourse is at risk. Given the financial pressures on for-profit news organizations and the temptation of chains to favour centrally produced news over local, we foresee a growing move toward a variety of not-for-profit and, possibly, charitable models that place a greater emphasis on their roots within given communities, which we describe more in the companion paper.
We thank our partners in this work, all of whom are dedicated to the proposition that news is a public good and that an absence of news removes an essential asset necessary for communities to know themselves and feel a sense of cohesion and common experience. Those partners are the Atkinson Foundation, Community Foundations of Canada, Communications Workers of America, the McConnell Foundation, Unifor, and the Vancouver Foundation. We also thank our research associate, Nordicity.
PPF’s 2017 report The Shattered Mirror: News, Democracy and Trust in the Digital Age documented how a huge revenue loss was sinking Canadian journalism. Newspapers across Canada were closing, leaving communities with reduced or no news coverage of civic affairs. To avoid closure, other news outlets were laying off staff, closing local offices, or merging them to create regional ones. From 2008 to 2016, 169 local media outlets closed and another 54 reduced services, a trend that accelerated in 2017, most notably with the swap of assets by Torstar and Postmedia.
Despite some limited startup activity, the trend lines are moving in a single and pronounced direction. While some digital-only publications have enriched the news ecosystem, so far they have tended to lack the journalistic intensity of the news media system of the previous century, perhaps because barriers to entry are low, global social media and search sites dominate national markets, and the necessary scale to prosper in the digital world is difficult to achieve in Canada. Therefore, organizations employing journalists and engaged in producing civic journalism aimed at broad audiences are most at risk.
The relevance of these findings goes well beyond those directly involved in the news industry, with major implications for Canadian politics and democracy. Access to quality news and information is critical in a democracy, as it enables citizens to know what their governments are doing, to hold those governments to account and to build a culture of civic participation and debate among citizens. Furthermore, access to quality local news is critical to a community’s civic health. The 2009 report by The Knight Commission on Information Needs of Communities in a Democracy, for example, found that “information is as vital to the healthy functioning of communities as clean air, safe streets, good schools and public health.”
The Shattered Mirror identified the factors that lead to reduced news coverage on a national scale and, as a result, we now have a good picture of what’s going on in Canada’s news ecosystem and why. The report also made 14 specific recommendations on what the Canadian government can do to reverse some of these trends and promote civic-based journalism in Canada. In the aftermath of the report’s January 2017 publication, PPF workshopped its recommendations with industry participants and ultimately chaired an effort supported by owners and unions of small and large newspapers and digital-only publications in English and French. This effort led to a second set of recommendations, directionally similar but less complex in policy terms, that was published under the auspices of News Media Canada. It called for reforms to the Canadian Periodical Fund that would provide government funding to encourage the employment of journalists and digital innovation.
Since 2017, the diminution of journalism has continued. There remain few examples of business innovation leading to marketplace success. The potentially bright spot is some experimentation in ownership models to finance journalism. Philanthropic groups, businesses, entrepreneurs and academics are starting to explore ways they can support, promote and contribute to journalism. We have also seen movement on the part of the Canadian government, and others, in this area. In the 2018 federal budget, the government said it would contribute $50 million over five years to support local journalism in underserved communities. It also tasked public servants with exploring “new models that enable private giving and philanthropic support for trusted, professional, non-profit journalism and local news.” It added that such models “could include new ways for Canadian newspapers to innovate and be recognized to receive charitable status for the non-profit provision of journalism, reflecting the public interest they serve.”
At the beginning of 2018, PPF decided to seek out some empirical evidence to show how journalistic coverage of communities across Canada has changed over the past decade. The Shattered Mirror showed the number of journalistic outlets being reduced by shrinking, merging or closing. The goal of this study is to pinpoint how these changes to outlets have affected changes to outputs. More specifically, it sought to answer the question: Is there less volume and depth of coverage of local communities now than there was 10 years ago? And, if so, what types of communities and coverage have been most affected?
To date, there has been virtually no known research of this kind in Canada; it is difficult to undertake, as we discovered, because of the limits to consistent databases of articles. We have not known which communities are more or less likely to have inadequate coverage, or what factors lead to this. We have not known what specific areas of coverage have been hit the hardest. Although hobbled by the paucity of high-quality database information, we worked with our research associate to design a study using keyword searches supplemented by human coding. As the federal government prepares its plans to distribute funds to support journalism in communities across Canada, and as other groups look at ways they might also assist local journalism, evidence of where gaps have opened becomes essential.
To complement this examination of the news coverage of local communities, we felt it imperative to look at the online media ecosystem, too—even if only anecdotally to start with. More detailed research and analysis will be required to answer the question of whether the online news media ecosystem, including news relayed via social media platforms, are adequately plugging the gap.
This report aims to answer the aforementioned question and, crucially, begin to plot a path to a more sustainable civic service journalism in Canada. It is not the end of these forms of inquiry; We would strongly recommend further detailed work of specific media ecosystems of the type undertaken by the Pew Charitable Trusts in the United States.
To answer the aforementioned questions, PPF, in partnership with our research associate, Nordicity, undertook a concerted, three-month research project that examined how the news coverage of democratic institutions and civic affairs in local communities across Canada changed from 2008 to 2017.
By democratic institutions, we refer to city halls, provincial courts and legislatures, local transportation authorities, school boards, police boards, and the like. We chose to focus on these specific institutions because they are critical to having an informed citizenry in a healthy democracy and they are more practical to define.
We selected the 20 communities on the basis of five key market characteristics, with the aim of ensuring a broad and representative sample. These were language (25 percent francophone or bilingual markets), region (each includes four newspapers), markets with closures (40 percent of markets), provincial capitals, and market size (small and medium only). We worked to ensure a balanced geographical spread, too.
We studied print news reporting of these democratic institutions by measuring the change in volume (measured in the number of articles) and depth (measured against five separate criteria: the presence of direct quotes; opposing perspectives; statistics or polling; historical context, and; illustrative examples). Specifically, we performed a quantitative analysis of the print reporting in all 20 communities selected, which included statistical and content analysis based on selected key indicators.
The full methodology is outlined at the end of this report.
Our empirical research indicates that the deteriorating state of traditional media has precipitated a drop in both the volume and depth of civic news reporting in local communities. The number of English-language newspaper articles published in the communities we examined decreased by almost half between 2008 and 2017, while the number of articles specifically reporting on civic affairs declined by more than a third. It also shows what has long been understood to be true: while critical to democracy, journalism about democratic institutions represents just a small portion of the total news offering.
We found that the number of civic affairs articles grew as a share of all articles, but this appears simply to be an artifact of other articles declining more rapidly than civic news. It may well be that publishers are trying to “protect the core” by concentrating limited resources on covering civic affairs at the expense of other topics, or that institutional coverage is easier to maintain under resource pressures. It could, as the research suggests, also stem from a cost-saving broadening of the geographic coverage footprint from local and regional to provincial in order to make better use of articles from external wire services, or internal services operating within newspaper chains themselves. Provincial news can travel from paper to paper, whereas local news tends to be less mobile, creating an incentive to cut local resources and pool provincial ones.
Our research also looked at where the gaps are emerging in terms of the types of civic affairs coverage. Specifically, this included civic reporting of three thematic areas: city hall, courts and legislatures. Consistent with our findings above, we found that the absolute number of articles covering each area had fallen between 2008 and 2017. In absolute terms, articles covering city hall declined by 38 percent, with declines of 30 percent and 29 percent in those reporting on courts and legislatures respectively. Our research revealed a slight decline in the share of articles covering city hall. While the share of articles reporting on courts and legislatures fluctuated over the decade, it recorded a slight increase between the start and end of the study. It is conceivable that this slight increase is due to the fact that legislatures and courts cover a wider geographical region and thus it may be the result of an increase in usage of newswire content, a recurring theme highlighted by our research.
We also broke down the types of civic coverage by the number of articles per newspaper edition. The trends here are very similar to those observed above. In short, the number of articles reporting on each of the three areas studied—city halls, courts and legislatures—have fallen by almost a half over the last decade.
Specifically, the number of articles across all three themes declined from 4.0 to 2.1 per edition—a drop of 48 percent—from 2008 to 2017. The number of articles reporting on city halls has fallen from roughly two articles per edition to one, while coverage of courts and legislatures has dropped from roughly one article per edition to one every other edition.
According to our research, not just the volume but also the depth of civic affairs reporting has declined over the last 10 years. Specifically, articles covering civic affairs showed declines in four of the five metrics we tested. This included the presence and level of analysis afforded to opposing views, historical context, statistics or polling and illustrative examples. That said, the scale of the decline was lower in comparison to volume and at least one example of each of our five depth indicators was present in the majority of the articles we studied. So, while in decline, the quality of coverage, as measured by our proxies, is eroding slower vis-à-vis the volume of coverage. Arguably, this is to be expected, as the slide in the number of journalists covering these issues is more likely to be felt first in the quantity rather than the quality of reporting, particularly if the reporting pool is biased toward those with more rather than less experience.
In addition to volume and depth, we also looked at the geographical scope of civic reporting, specifically what level of civic affairs—neighbourhood, municipal, regional, provincial or national—articles were covering. It follows that given the state of decline in traditional media, we expected to see an increase in the proportion of national and provincial coverage as the number of journalists and, in turn, the number of local-level articles fall, with resulting gaps backfilled, as mentioned, with more newswire reporting. The research showed that the proportion of provincial-level coverage has indeed increased over the last decade at the expense of coverage at municipal and regional levels. More analysis needs to be done to measure to what extent this increase is the result of wire service or staff articles. In any case, we seem to be witnessing two clear trends: fewer articles and a greater proportion of articles from outside the community itself.
As part of our study, we looked at how print reporting of civic affairs has changed across different communities in Canada from 2008 to 2017. These communities varied in size, language, and whether or not they experienced a newspaper closure in recent years. We found that all these communities would appear to have been affected by the deteriorating state of traditional news media. All have recorded a drop in the volume and depth of civic affairs reporting over the last decade. This includes both small and medium-sized communities (e.g. under 100,000 and 600,000 inhabitants respectively) and both anglophone and francophone ones. Significantly, communities that have not experienced a newspaper closure recently have seen a decline in civic affairs reporting, too. In addition, all have seen the proportion of civic affairs coverage at a provincial level grow at the expense of those at municipal or regional levels.
Where communities have differed has been the scale of the impact. Unsurprisingly, the starkest contrast has been between communities with and without newspaper closures. Those with closures have seen a bigger fall in both volume and depth of civic affairs reporting. Communities with closures have experienced a greater growth in provincial coverage, at the expense of regional and municipal coverage, than those without. It is axiomatic that when a community loses its newspaper, it loses the news contained within that newspaper. That said, the illustration of that point is stark.
Our research also appears to indicate that small and medium-sized communities have been impacted differently. For example, smaller communities have seen a sharper decline in the volume and depth of reporting over the last decade. But the overall impact has arguably been greater in medium-sized ones as coverage of civic affairs has been consistently lower over the last decade. So, arguably, declines have been more acutely felt. Medium-sized communities have also witnessed more growth in provincial coverage, which has almost entirely replaced reporting from a regional perspective.
The scale of decline in the volume of civic affairs coverage appears to have differed remarkably in anglophone and francophone communities, according to our research. For example, the fall in the number of civic affairs articles appears to have been more acutely felt in English markets than bilingual ones. As regards the depth of reporting, the picture seems more balanced. That said, as with small and medium-sized markets, one could argue that the impact has been greater on francophone ones as the depth of coverage—as per the five indicators we tested against—has consistently been lower over the last 10 years. Similarly, our research found that while both English- and French-language articles had fallen in depth, the impact may be greater on the latter as, again, for the same reason.
Similarly, not all markets felt the decline equally in terms of the types of civic affairs coverage. In general, markets that experienced a closure tended to have lower levels of city hall reporting than those without. In small markets, the share of reporting on city hall has remained stable at around 50 percent since the initial drop at the start of the decade, whereas the drop in medium-sized markets was accompanied by a decline in provincial-level reporting (legislatures and courts), relative to municipal (city hall) coverage. This would appear to support the idea set out earlier in this report, namely that news outlets are “protecting the core”. Finally, the research appears to reveal a stark contrast between English markets and English-language coverage in bilingual markets. Not only has the decline in the volume of coverage been most dramatic in English markets, the coverage of city halls has steadily lost ground to legislatures and courts. Bilingual markets have shown a modest increase in coverage of city hall against a slight decline in provincial-level (i.e. courts and legislature) reporting.
The picture appears to be mixed in terms of how the number of articles covering city hall, courts and legislatures per edition has evolved over the last decade. For example, markets with and without recent closures have experienced a similar decline in the number of articles reporting on city hall, court and legislatures. In contrast, small and medium-sized markets have diverged. While both have experienced dips in all three types of reporting, the scale of the decline as regards the coverage of courts and legislatures has been considerably more pronounced in medium-sized markets. Anglophone and francophone markets have differed even more. While anglophone markets have seen a decline in all three subjects of reporting, francophone markets have, in contrast, seen a slight increase in the coverage of city hall. Further, coverage of courts in francophone communities appears to have been unaffected.
This research validates the concern that the weakened economic state of the print news industry is leading to less local and community coverage in general and less civic coverage in particular. The information that reporter-based news organizations provide to inform democratic choice, hold officials to account, and allow communities to know themselves better is in sharp decline.
However, it is possible that civic coverage is still occurring through a number of other means. For example, many civic bodies now post minutes of meetings on their websites. In some cases, live video streams of civic events, such as municipal meetings, are available. There are also numerous examples where residents have created social media discussion groups to share and debate issues affecting their local communities, including the impact of municipal decisions. While new digital sources of civic and local news coverage may be emerging, the role of good journalism in analyzing information, bringing context to the discussion and generating new insights should not be underestimated.
Government should have no interest in bailing out the news industry; its interest is in ensuring democracy is well-served by having a robust means of specifically informing citizens of civic activities in their communities and, more generally, of being connected to one another through common bases of information. The print news industry has played a vital role in this area in the past and there may be significant consequences for democracy from its decline.
The following summarizes the methodology for the statistical and content analysis performed by Nordicity on print newspaper coverage in 20 Canadian markets over the 10-year period between 2008 and 2017.
In consultation with PPF, Nordicity identified 20 communities across Canada that provide a sampling of characteristics hypothesized to shape the local news ecosystem. Figure A shows the distribution of markets by population size (small communities are defined as having a population under 100,000 while medium are under 600,000), region and language (red text indicates markets not covered by Infomart, an online article database for English-language newspapers). It also indicates which markets are provincial/territorial capitals and in which markets newspaper closures occurred during the study period.
Nordicity worked with PPF to develop 13 keywords that would occur in articles discussing civic institutions. These keywords were used as search terms to select articles relevant to these issues and are summarized in figure B.
The statistical analysis relies on search results exported to Microsoft Excel from Infomart. The data covers the period from 2008 to 2017 (inclusive), and represents all articles that were available in digital form for all newspapers in each market.
The total number of articles (813,122) is based on a search performed for all articles containing the keyword “a” over the period from 2008 to 2017. (Infomart’s search interface does not permit searches with no search terms; “a” was deemed the most inclusive approach to obtain results representing articles.) Civic articles referenced in this analysis (47,723 in total) include all articles that were retrieved using the 13 search terms specified in the methodology.
Some publications appear to have started digitizing articles during the study period (2008-2017). As such, we recognize that article counts may be artificially low at the beginning of the study period. This study observed a significant decline in articles (despite facing a headwind), so we note that declines are likely understated.
Figure C shows the total number of articles associated with each market in the Infomart data set that was used to conduct the statistical analysis. The number of civic articles is shown in brackets after the total.
Nordicity and PPF designed a more targeted list of search terms to distinguish between coverage of three reporting themes: city hall; courts; and provincial legislatures. Figure D summarizes the search terms that were associated with each theme. In contrast to the generalized ‘civic’ articles, this list excludes terms that were found to have cyclical tendencies (particularly election coverage), and would therefore obscure the visibility of trends in the results. Because this analysis was limited to Infomart’s English-language data, only English search terms were used. Note that the 13 search terms used to define civic articles was also used for the content analysis, which was conducted for both English and French content.
The analysis is based on a sample of articles collected from 20 markets in both English and French. Articles in the sample were randomly selected. The selection process involved retrieving a random article for each search term while cycling through each of the 13 search terms until the quota was met for each market. If there were no remaining articles for a particular search term, the process continued to the next search term.
For markets that were not available in the Infomart database, researchers performed a manual version of the search using the online search functions of the newspapers’ websites. While functionality of online searches varied from newspaper to newspaper, researchers exercised judgment to ensure the collection of a uniform sample for the beginning and end period within each market.
Several markets did not have digitized articles available as early as 2008. For these markets, the research team selected articles from 2012 to perform the comparison over time. In the case of two French-language markets (Rimouski and Timmins) data was available only for 2017.
For articles covered by Infomart, 40 civic articles were coded for each year. Some markets had fewer than the 40-article threshold available. In these cases, the research team coded all available articles in the intended year. For markets not covered by Infomart, 20 articles were coded for each year. In cases where no articles were digitally available in 2008, the “before” sample was taken from 2012. Figure E shows the breakdown of articles coded by market.
 It’s worth adding that this decline is likely to be understated as some publications only started digitizing articles during the study period. So, article counts at the start of the study period may be artificially low.
 The other metric was direct quotes, which recorded a slight increase.
 A word of caution, however. Our sample of francophone communities and French-language content was smaller (25 percent) compared to anglophone and English-language content. So, a larger sampling would be required to substantiate this observation and the other points raised in this paragraph.