Abstract
Sustainability has moved from fringe topic to headline news and key policy discourse in its own right. Yet, the sustainability discourse remains fragmented, with a diverse set of challenges receiving vastly different levels of attention. Nevertheless, the vast majority of previous studies have focused on media attention to climate change, whereas other sustainability challenges have received much less attention in the academic literature. In this paper, we explore trends and patterns in media coverage across a set of ten sustainability challenges. In particular, we are interested in the extent to which the recent trends and patterns in coverage that have been well-documented for climate change are reflected by other sustainability challenges. We utilise a sample of 23 broadsheet newspapers from five different countries (Australia, Canada, Germany, UK, US), covering a 17-year period from 2000 to 2016. Using the agenda-setting literature as a starting-point for our enquiry, we then turn to the toolset provided by financial econometrics to develop a basic typology of media attention focusing on the two dimensions information/noise and seasonality/non-seasonality. We find that media coverage on climate change, poverty and HIV/AIDS can mainly be characterized as information, whereas the remaining seven issues included in our study appear noise-driven. Seasonal patterns in coverage appear most pronounced for socioeconomic issues. Media attention to biodiversity and cleaner technologies has been crowded in by increased coverage on climate change. At the same time, we find clear divergences from overall trends and patterns at the level of different countries and individual newspapers.
Introduction
Sustainable development has moved from fringe topic to headline news and key policy discourse receiving unprecedented public attention. This is good news given that media coverage and related public concern are seen as important preconditions for effective policy responses (Baumgartner and Jones, 1991; Dearing and Rogers, 1996; Downs, 1972). Whilst no consensus exists regarding the exact nature of the relationship between the media agenda and the public as well as policy agendas, it has long been acknowledged that the mass media clearly have an important role to play in directing public attention towards specific policy concerns (Cohen, 1963; Neuman, 1990).
Conceptually, sustainable development covers a wide range of natural and socioeconomic challenges (WCED, 1987). However, these challenges receive very different levels of attention and reflect fundamentally different trajectories. For example, whilst climate change has received unprecedented levels of public attention (Barkemeyer et al., 2017; Boykoff, 2007; Boykoff and Roberts, 2007), media coverage on sustainability challenges such as biodiversity or soil erosion has been much more limited (Barkemeyer et al., 2013). Likewise, academic attention in this context has focused almost exclusively on climate change (for overviews see e.g. Boykoff and Roberts, 2007; Schmidt et al., 2013). This raises the question whether the wealth of findings that has been generated in the context of climate change also holds for other sustainability challenges.
In this paper, we break the sustainability discourse down to ten specific sustainability challenges and we explore their media coverage. In particular, we are interested whether the recent trends and patterns in coverage that have been well-documented for climate change are also reflected in coverage of other environmental and socioeconomic challenges. We utilise a sample of 23 broadsheet newspapers from Australia, Canada, Germany, the UK and the US to map out sustainability-related media attention between 2000 and 2016. Using the agenda-setting literature as a starting-point, we then turn to financial econometrics to develop a basic typology of media attention. We focus on two distinct but inter-related dimensions, information/noise and seasonality/non-seasonality. On this basis, we explore commonalities and differences in sustainability-related media coverage between climate change and other sustainability issues, thereby going beyond the dominant focus on climate change. Our large-scale empirical analysis helps us to get a more nuanced understanding of the role of the media in relation to sustainability agendas and public concern, as well as resultant implications for the policy process.
The results show markedly different coverage profiles for different sustainability challenges as well as a considerable amount of newspaper-level variation in coverage, pointing to distinctly different underlying dynamics. The identification of these trends and patterns allows us to move beyond the simplistic assumption that sustainability as such has become – and will continue to become – an increasingly important public concern over time and thus inevitably trigger suitable policy responses. Only coverage on climate change, poverty and HIV/AIDS is classified as information, whereas coverage on the remaining seven issues is driven by noise. Poverty and HIV/AIDS are also the two issues that show the clearest seasonal patterns in coverage, peaking in the run-up to Christmas and World AIDS Day, respectively. In addition, we identify interaction between different sustainability issues: whilst coverage on biodiversity and cleaner technologies seem to have been crowded in by the dramatic increase in media attention to climate change, this appears to have had a negative impact on coverage on socioeconomic challenges such as HIV/AIDS, human rights or labour rights.
The remainder of this paper is structured as follows. The next section provides a brief overview of the emergence of sustainable development as a major policy discourse. Next, the role of the mass media in promoting as well as shaping the sustainability-related public agenda and policy agenda is reviewed, before we develop a classification of issue-level media coverage that is grounded in financial econometrics. Then, the methods applied in this study are described and justified, followed by the presentation and discussion of findings. We conclude by sketching out potential limitations to the research methods applied in this study as well as presenting avenues for future research.
Sustainable development and the mass media
The emergence of sustainable development
The origins of sustainable development date back several centuries. For example, Malthus (1798) identified one of today’s key elements of sustainable development, the link between overpopulation and the emergence of social problems, as early as the late 18th-century. As a generally accepted paradigm and policy discourse, sustainable development is a much more recent phenomenon that started to gain prominence in recent decades. Today, sustainable development has gained widespread political and public authority and has arguably become “the common currency of almost all players in the environmental arena” (Jacobs, 1999). Definitions of sustainable development typically refer to the work of the Brundtland Commission and their ground-breaking report “Our Common Future” (WCED, 1987). In essence, the Brundtland Commission defined sustainable development as “development that meets the needs of the present without compromising the ability of future generations to meet their own needs” (WCED, 1987: 43).
In line with the mandate of the Brundtland Commission, i.e. to address environmental issues as well as problems that traditionally have been considered the domain of international development, this definition captures both international development and environmental protection objectives (Robinson, 2004). Ultimately, sustainable development therefore covers an astonishing range of very different environmental and socioeconomic challenges, some of which are global in nature and some of which are more context-specific. In practice, not all of these challenges may play a prominent role in the discourse on sustainable development. In addition, it has repeatedly been pointed out that the discourse on sustainable development should in fact be seen as a range of different framings and multiple discourses, some of which even appear to be competing or even mutually exclusive (Bulkeley et al., 2013; Redclift, 2005). Irrespective of these difficulties to arrive at commonly accepted operationalisations of sustainable development, it is nevertheless possible to identify specific sustainability challenges according to key normative agendas and policy initiatives such as the UN Sustainable Development Goals (SDGs). The UN SDGs – in line with the Brundtland definition – reflect a Triple Bottom Line orientation (Elkington, 1997), addressing a wide range of economic, social and environmental challenges.
Sustainability challenges considered in this study.
Sustainable development and the media agenda
The mass media inevitably play an important role in the move towards sustainable development. Norris (2000) assigns three distinct functions to the media in a representative democracy, i.e. acting as a civic forum; a mobilizing agent; and a watchdog. In the context of sustainable development, examples of the key role of the media include the mobilization of the general public in the wake of large-scale incidents such as the Bhopal disaster, the Enron scandal, or the Nike sweatshop labour scandal. Beyond specific incidents, the media also play an essential role in gradually building up public attention to challenges such as climate change, biodiversity or gender equality. The media represent a “mediated public sphere” that shapes global citizenship and may play a powerful role in enhancing a sense of global connectedness (Szerszyski et al., 2000), thus acting as a mobilizing agent.
Media communication takes a number of different forms and channels, ranging from printed newspapers to televised news and the use of social media. It has long been suggested that print publications have a stronger agenda-setting impact than for example television news (Boykoff, 2010; McCombs, 1977). Furthermore, different types of print publications can be expected to trigger different impacts. In the UK context, broadsheet newspapers such as The Guardian, The Independent or The Times can be expected to provide the highest amount of coverage on political issues and also to have the highest agenda-setting impact for policy makers and the general public (Carvalho and Burgess, 2005; Sparks, 1987). In addition, they have been argued to serve as “inter media agenda-setters” (McCombs, 2004), and thus to influence coverage of tabloid newspapers and other news media.
The agenda-setting hypothesis was originally formulated by Cohen (1963: 13) who suggested that the media “may not be successful much of the time in telling people what to think, but [they are] stunningly successful in telling its readers what to think about”. In essence, it captures the relationship between media attention and public attention. A range of models have since been developed that theorise the link between media coverage, public opinion and public policy (Baumgartner and Jones, 1991, 1993; Dearing and Rogers, 1996; Downs, 1972; Kingdon, 1984; Neuman, 1990). Downs (1972) suggests that different characteristics of an issue can be expected to mediate the impact of media coverage on the public and policy agendas. Likewise, a range of factors have been found to influence the magnitude (and direction) of the relationship between the media agenda and public and policy agendas, including specific characteristics of the issue that is reported on (Neuman, 1990), the tone and framing of media coverage (Carvalho, 2005; Carvalho and Burgess, 2005), or characteristics of the recipients of information provided by the media (McCombs, 2004; Reynolds et al., 2010; Spence et al., 2012). All of these factors influence media responses to a given real-life event.
At a more general level, however, a robust link has long been identified between media attention and public perceptions of the importance of an issue (Atwater et al., 1985). Hence, the volume of media coverage has frequently been used as a proxy for the level of public attention (e.g. Carvalho, 2005; Dearing and Rogers, 1996; Mueller, 1973). Therefore, using media attention – measured by the volume or frequency of media coverage on a given issue – has become a widespread approach in agenda-setting research (Cook and Skogan, 1990; Howlett, 1997; for an overview and critique see e.g. Kiousis, 2004). Examples in the context of sustainability include media coverage on air pollution (Forsyth, 2014), HIV/AIDS (Dearing, 1989), climate change (Boykoff, 2007; Brossard et al., 2004; Schmidt et al., 2013) or poverty (Bunis et al., 1996).
However, it is important to note that even though a substantial body of literature has recently started to explore the role of mass media in sustainability, these enquiries have almost exclusively been restricted to climate change. Beyond the large number of studies examining volumes of media coverage on climate change in various geographical contexts (for an overview see Schmidt et al., 2013) and for example enquiring whether trends in media coverage on climate change reflect a tipping-point (Russill and Nyssa, 2009) or punctuated equilibrium (Holt and Barkemeyer, 2012), recent research has explored the role of TV coverage (Gavin and Marshall, 2011) or social media such as Twitter (Jang and Hart, 2015; Kirilenko and Stepchenkova, 2014), and has analysed the role of the mass media in relation to individual policy events (Freudenburg and Muselli, 2010), or in relation to specific climate-related problems such as sea level rise (Yusuf et al., 2015). In stark contrast, the role of the mass media in the context of sustainability challenges other than climate change has clearly remained under-researched.
Analytical framework
We explore trends and patterns in media coverage across a set of ten sustainability challenges. We are interested in the extent to which trends and patterns in coverage of climate change (for overviews see e.g. Boykoff and Roberts, 2007; Schmidt et al., 2013) are also reflected at the level of other environmental and socioeconomic problems. Absolute coverage levels are a first, basic way to describe media coverage. Higher coverage can be expected to result in higher levels of public attention and more prominent position on the public and policy agenda. Beyond absolute coverage levels, however, media attention can also reflect specific trends and patterns over time. We will explore these trends and patterns using the toolset provided by financial econometrics as our analytical lens. In particular, we will focus on the concepts of information/noise and seasonality/non-seasonality, respectively.
Information versus noise
Drawing from the finance literature, a useful distinction can be made between “noise” and “information” (Black, 1986): here, noise can be seen as a potentially random or even irrational element that distorts otherwise efficient markets. Information then represents the opposite of noise, leading to accurate and rational decision-making. Applied to media attention, this distinction helps us to contrast (a) coverage on a given issue that merely appears as background noise with no apparent links with previous or subsequent coverage on this issue (i.e. noise), with (b) coverage that is part of an overarching thread or storyline, for example coverage on a specific event that triggers feedback from policymakers or other societal actors (i.e. information). Whilst media coverage dominated by noise would reflect a somewhat flat trend, media coverage that is purely characterised by information is time-dependent and could to a certain extent be explained by previous coverage levels. One such “information” pattern can be described as issue-attention cycle (Downs, 1972), with peaks in coverage typically alternating with periods of relative calm. Peaks in media coverage are followed shortly afterwards by saturation or boredom effects (Downs, 1972; Neuman, 1990). Media coverage and public attention reach a plateau and subsequently return to previous levels.
As “with time series data, it is the dependence between values of the series that is important to measure” (Shumway and Stoffer, 2013: 57), the concepts of noise and information are related to stationarity. Stationarity, as defined (in its weak form) by Box et al. (2015: 07), refers to a time-series that “remains in statistical equilibrium with probabilistic properties that do not change over time, in particular varying about a fixed mean level and with constant variance”. The distinction between stationarity and non-stationarity also applies to media coverage time series. As a stationary process, the level of media coverage is supposed to be mean-reverting, i.e. fluctuating around a constant level. This weak time-dependency means that there is no change in media attention to a given issue in the long run. Instead, we merely observe random variation in the levels of media coverage, with only a temporary impact on the level but not the trend (noise). Such a process is said to have a “limited memory” or “no memory”. In other words, there are no changes in the communication pattern over time.
The higher the degree of information in relation to noise, the more likely that we can identify non-stationarity in a time-series. When modelled as a non-stationary process, the level of media coverage has a stronger time-dependency, since the average level of media coverage (trend) is time-varying. The long-standing debate introduced by Nelson and Plosser (1982) suggests two sources of non-stationarity and, as a consequence, two classes of non-stationary models: (1) the first class “consists of those that can be expressed as a deterministic function of time, called a [deterministic] trend, plus a stationary stochastic process with mean zero” (Nelson and Plosser, 1982: 141). Such a process is assumed to be trend-reverting and suggests, here, a global and continuing trend in levels of media coverage and the communication pattern: an increase (decrease) in media attention to a specific issue, reflecting increasing (decreasing) awareness of the issue over time. (2) The second class includes purely stochastic processes with a time-varying mean but no deterministic trend. The mean level varies randomly (stochastic trend): random changes in the level and the slope of the series. Such series never recover from shocks (unexpected variations) and are the most sensitive to them. Shocks have a “permanent” effect in the sense that every shock contributes to the trend. Hence, series are said to have a “perfect memory”. The extreme case is a Random Walk process and other cases can be described as randomly-varying processes. Even if they have a non-constant but unpredictable trend, they can typically have long periods of upward or downward trends with sudden and unpredictable changes in direction (Hyndman and Athanasopoulos, 2014).
In the context of media coverage and public opinion, these lasting impacts of shocks have been theorised as “punctuated equilibria” (Baumgartner and Jones, 1991, 1993). Here, a triggering event changes the response function of the media and the general public: media attention does not drop back to the levels identified immediately before the given focusing event, but prolonged levels of above-average media attention can be identified. Examples in this context might be a series of high-level events such as the Nobel Peace Prize awarded to the IPCC, moving climate change into the spotlight of public attention (Holt and Barkemeyer, 2012); or the Fukushima disaster, creating a watershed for the anti-nuclear movements in countries such as Germany or Italy (Butler et al., 2011). At the same time, the opposite effect might also be conceivable: coverage on a given issue might be crowded out for a prolonged period of time based on an unrelated focusing event that shifts media attention away from a specific problem, and thus creating a negative punctuated equilibrium.
Saturation and boredom effects might result in a more subtle decline in media attention. Here, media coverage on a given issue may alter its public response function over time, reflecting ritualization or routinization effects. Both the media and the general public might quite simply become accustomed to events of a certain type such as terrorist attacks (Roy and Ross, 2011), violent conflicts (Volkmer, 2008), or homelessness and famine (Bunis et al., 1996). Likewise, the characteristics of the issue at hand may change over time, again leading to an alteration in public response to real-life cues and related media coverage. People might find it more difficult to relate to a given issue after a sustained period of relatively high media coverage. Geographical proximity to the issue (Berkowitz and Beach, 1993) may also play an important role in this context.
Seasonality versus non-seasonality
Whilst the analysis of a trend component dealt with the relationships within periods, the analysis of a seasonal component deals with relationships between the observations for the same period in successive years. Seasonality describes a recurring pattern in media coverage. Granger (1979) suggests four classes of causes of seasonal fluctuations in economic data that may be also correlated: (i) Calendar, because of the timing of public holidays and other common events; (ii) Timing decisions, because some decisions made by individuals and institutions are inclined to occur at similar time each year, and are generally deterministic and preannounced; (iii) Weather, as a natural cause of the true seasonality which have direct effects; (iv) Expectations, since “they can lead to plans that then ensure seasonality”. People may expect a seasonal pattern that might however vanish or change as more information becomes available. Beaulieu and Miron (1990) find that a Christmas shift in preferences and synergies across agents (increasing returns or other synergies, rather than shift in technologies) are the key determinants of seasonal patterns around the world.
Bunis et al. (1996) studied seasonal variation of attention to homelessness and famine and of sympathy and goodwill toward the homeless, finding a link between seasonality and shifting preferences. Hilgartner and Bosk (1988) point to an “institutional rhythm” and a culturally based temporal ritualization. A limited “carrying capacity” for social problems might limit the number of problems that gain widespread attention at one time. A specific societal challenge therefore competes with other challenges for attention in the public arena, and may only at certain points during the year be able to attract significant levels of attention. The time series literature distinguishes different reactions to seasonal shocks (Hylleberg et al., 1990; Parzen and Pagano, 1979; Pierce, 1979). However, for the purposes of this paper, we focus more generally on whether or not a time-series reflects any type of seasonal pattern.
Conceptual matrix.
Method
The method consists of a three-stage analysis involving data collection; an automatic modelling process leading to the identification of best-fitting models depending on their trend and seasonal components; and a classification based on the concepts information/noise and seasonality/non-seasonality. We explore 230 individual time series across the set of 10 sustainability issues and 23 newspapers.
Data collection
Data was collected using keyword searches of the LexisNexis and Factiva newspaper repositories. For each of the ten sustainability challenges, stemming algorithms were developed that allowed for the inclusion of possible variations in spelling, conjunction or misspelling of key terms. These were developed for the languages included in the sample (English and German). All articles that contained at least one mention of the respective search term were counted.
A text mining routine converted keywords into frequency tables, displaying the raw frequencies in terms of the total number of articles for each newspaper and month during the review period. It is important to note that this analysis focuses levels of media attention rather than the positions adopted towards this issue: the analysis identifies coverage levels and therefore the extent to which – but not the way in which – an issue has been discussed in the public arena.
The selection criteria for the sampling of newspapers included circulation, area of circulation and, political alignment. Where possible, we aimed to create balanced country subsamples of newspapers that included both centre-left and centre-right positions. Newspapers were selected that were not predominantly local or regional in scope and therefore, at least to a certain extent, reflected their national public agendas (detailed in Appendix 1). Priority was given to national broadsheet papers as these can be expected to provide the highest amount of coverage on political issues and to have the highest agenda-setting impact for policy makers and the general public (Carvalho and Burgess, 2005; Sparks, 1987).
Automatic modelling of time series
Our time series describe levels of media coverage, expressed as the monthly number of articles related to a specific issue published over 204 months. Their dynamics are captured by identifying the best forecasting model, i.e. the model that best fits the data, in the spirit of the Box–Jenkins methodology (see e.g. Enders, 1990; Shumway and Stoffer, 2013; Tsay, 2005): each model explains the level of media coverage at the time t on the basis of past values of the same variable and/or by past forecasting errors. 1 The automatic modelling process developed in SPSS Forecasting (Expert Modeler) identified the models. Automatic modelling is considered as performing as competent modellers (see De Gooijer and Hyndman, 2006 for references). Crucially, for our purposes, it ensures unbiased results based on common measures for the goodness of fit. 2
Previous studies have applied the Box–Jenkins methodology and similar processes to explore media attention: Schäfer et al. (2014) applied ARIMA (Auto Regressive Integrated Moving Average) modelling as a part of their methodology, to explore the drivers of media attention of climate change in Australia, Germany and India from 1996 to 2010. Liu et al. (2011) applied a VAR (Vector Auto Regression) approach to examine the linear interdependencies among time series, from 1969 to 2005, for US media attention and congressional attention by including climate problem indicators, high-profile international events, and climate science feedback. Our modelling approach differs from the latter by focusing on univariate time series. Furthermore, it also differs from the former in following four ways: (1) by exploring both ARIMA models and Exponential Smoothing (ES) models (see Gardner, 1985, for a review of ES models) as studied in the time series literature; (2) by modelling seasonality of time series, if any, either with Seasonal-ARIMA (Box and Jenkins, 1968) models, pure seasonal models and any other ES models including a seasonal component; (3) by allowing the detection of outliers, as Additive or Innovative Outliers (Chang et al., 1988; Fox, 1972; Peña, 1984) or Level Shift and Transient Outliers (Box and Tiao, 1975; Tsay, 1988), that may be the result of unusual external events, impacting the time series in a permanent or a transitory way, that can even affect the specification of ARIMA and S-ARIMA models (Box et al., 2015); and (4) by exploring additional issues and a larger group of countries and newspapers at the level of individual newspapers (Appendix 2).
Classification of best-fitting models
Classification of models that best-fitted the time-series data. a A (0,0,0)x(0,0,0) model describes a constant (mean) model with white noise. A (p,d,q) model is an ARIMA model with an autoregressive component of order p, a moving-average component of order q and d order of differencing (here, d = [0,1]). A (0,0,0)x(P,D,Q) model represents a non-multiplicative Seasonal-ARIMA model, with P, the seasonal order of the autoregressive part, Q, the seasonal order of the moving-average part and D, the order of seasonal differencing (here, D = [0,1]). A (p, d, q)x(P, D, Q) model is a multiplicative Seasonal-ARIMA, with a non-seasonal (regular) component and a seasonal component. C represents the (non-zero) constant term, if any. The constant term, C, is affected by the regular and the seasonal parts of the model. For a non-multiplicative seasonal model, C is the seasonal intercept.
A non-zero mean supposes that there is no trend: neither growth nor decline in coverage as illustrated by a horizontal line. In some specific cases, the stationary process can converge to a zero (constant) mean.
To illustrate this initial classification incorporating different trend and seasonal components, Figure 1 presents a number of representative examples of different types of time-series taken from the results of the automatic modelling process. Shown are the original time-series expressed as monthly number of articles on a given issue (red line); the fit line, representing the forecasting model that best fits the data (blue line); and the upper and lower confidence limits for the predicted values (dotted lines).
Illustrative examples of different types of time-series. (a) An example of a stationary mean reverting time-series. Here, the best fitting model is an ARIMA (0,0,0)x(0,0,0) + positive constant (4.781 articles published in average) resulting in a constant + white noise process. We see an innovational outlier in September 2010, an additive outliers in April 2011 and a seasonal additive outlier in September 2011. (b) The role of level-shift outliers on a stationary time-series. We observe two significant level shifts in May 2010 (−0.361) and February 2016 (+1.884) coverage, with subsequent coverage fluctuating around a different average. (c) A deterministic (non-stationary) trend. The model is an ARIMA (0,1,1) + a negative constant (−0.282). Without an AR coefficient, the slope is −0.282, i.e. the level of media coverage declined by 0.282 articles per month between January 2000 and December 2016. (d) A stochastic (non-stationary) time-series. The model, describing the logarithm of the number of articles published, is an ARIMA (0,1,1) with a MA coefficient of 0.712. (e) A stationary series with a seasonal component. The logarithm of the number of articles published is modelled as an ARIMA (1,0,0)x(1,0,1)+ positive constant (3.2). (f) A non-stationary series with a seasonal component. The model is an ARIMA (0,1,1) × (1,0,0).
Results
Minimum monthly coverage/Maximum monthly coverage/Total cumulated (2000–2016).
At the same time, clear country-level deviations from these overall patterns can be observed. Peaks in coverage related to climate change appear particularly pronounced in all Australian as well as the majority of UK-based and Canadian newspapers included in the sample. On the other hand, coverage on human rights is less pronounced in Australian newspapers, which show a clear emphasis of corruption-related coverage. Labour rights receive relatively high coverage levels in German newspapers, whilst the same applies to biodiversity in Australian an UK-based publications.
Co-occurrences of key terms (pairwise comparisons, entire sample, 2000–2016). This table presents co-occurrences of key terms across the entire sample. For example, throughout the review period climate change was mentioned in 24% of all articles mentioning biodiversity (2nd column, 4th row). At the same time, articles mentioning biodiversity accounted for 3% of total coverage on climate change (4th column, 2nd row). Light grey indicates values >5%; dark grey indicates values >10%.
Issue-level patterns in coverage across the set of 23 newspapers are described in Figure 2. In all ten cases, clear variations in coverage can be identified. Examples are the sudden increase observed in coverage on poverty and climate change, with the “Make Poverty History” campaign in 2005 and a series of climate change-related events between 2007 and 2009 triggering prolonged periods of high coverage, respectively. On the other hand, two clear peaks can be identified in coverage on malaria, but here coverage immediately drops back to levels that had been observed prior to these peaks. In order to better understand the processes underlying variation of coverage on the ten different sustainability challenges, we focus on the level of individual newspapers.
Issue-level coverage (overall sample, January 2000–December 2016).
Classification of issues (newspaper level).
A general pattern emerging from Table 6 is the considerable diversity in newspaper-level coverage underlying the aggregate issue-level trends observed in Figure 2. To varying degrees, however, commonalities can be identified in coverage across the set of 23 newspapers. Climate change represents on end of the spectrum: here, 19 out of 23 newspapers (86%) are classified as information. Malaria represents the other extreme, with clearly noise-driven coverage and only two out of 23 newspapers (9%) classified as information. For the other eight issues, a more mixed picture emerges. Seven out of 10 issues are dominated by noise; poverty and HIV/AIDS are the only other two issues for which the majority of newspapers are classified as information (13 and 12 out of 23 newspapers, respectively). At the same time, poverty and HIV/AIDS are also the only two issues in our sample for which the majority of newspapers show clear seasonal cycles in coverage. Coverage on HIV/AIDS tends to peak in the run-up to World AIDS Day on 01 December, whereas poverty-related coverage tends to surge during the Christmas period. In more general terms, seasonal patterns appear notably more pronounced for socioeconomic issues when compared to the environmental issues included in our sample. Considering all issues, 43% of all newspaper-issue combinations are classified as seasonal.
It should be noted that the diversity of patterns observed above is not randomly distributed across newspapers. Instead, a number of country-level patterns emerge from the best-fitting models presented in Appendix 2. For example, whilst the overarching picture for biodiversity is that of noise, three out of six UK newspapers are classified as information in relation to this issue. Along similar lines, cleaner technologies are typically classified in Canadian and US-based newspapers. The US also stand out with regard to climate change, where two out of five newspapers are classified as noise. At the same time, individual newspaper characteristics also appear to play a role in shaping patterns in coverage. This can be illustrated by two UK-based broadsheets, The Guardian and The Times. In The Guardian – which has long held a reputation for comprehensive coverage on sustainability-related issues – eight out of ten sustainability issues are classified as information. In stark contrast, coverage in The Times clearly appears noise-driven: here, only three out of ten sustainability issues are classified as information, and all issues except labour rights show clear seasonal patterns in coverage.
Discussion
Media coverage can be expected to influence levels of public attention, which in turn has the potential to shape policy agenda. Levels of media coverage are therefore an important precondition for policy activity (Crow et al., 2016; Yusuf et al., 2015). Our analysis has uncovered a multiplicity of patterns in sustainability-related coverage. More specifically, the distinction between information and noise (Black, 1986) provides a useful theoretical lens to explore similarities and differences across media coverage on the set of ten sustainability challenges. Crucially, the multitude of dynamics identified in media coverage on different types of sustainability challenges shows that the wealth of findings that has recently been generated in the context of climate change (e.g. Boykoff, 2007; Boykoff and Roberts, 2007; Schmidt et al., 2013) does not appear to be generalizable to other sustainability challenges.
The overarching picture across our sample is one of noise. Only for three issues (climate change, poverty and HIV/AIDS), the majority of newspaper-level time-series reflect information in the form of deterministic or stochastic trends. However, for all issues other than malaria – which is classified as noise in almost all 23 newspaper-level time-series – we can identify at least some newspapers in which issue-level coverage can be classified as information. Only in very few cases, we can observe deterministic trends that would also allow us to forecast coverage on a given sustainability challenge. In the vast majority of cases, stochastic trends can be identified, with (relatively few) “information peaks” that drive level shifts in coverage. Here, in line with the idea of “punctuated equilibria” (Baumgartner and Jones, 1991, 1993), specific events clearly impact subsequent media attention to a given sustainability challenge. For climate change, positive level shifts can be identified at different points in time, most notably linked to the 2009 UN Climate Change Conference.
Beyond climate change, poverty and HIV/AIDS are the only other issues in which the majority of time-series reflect information rather than noise. On the one hand, this is good news in terms of reflecting identifiable threads and storylines in coverage, thus creating windows of opportunity for policy change. On the other hand, coverage on both poverty and HIV/AIDS appears to be dominated by recurring patterns in coverage over time. Here, the influence of journalistic work and editorial decision-making, rather than real-life events, appears to be particularly pronounced. Ritualization of media coverage has previously been discussed as a risk linked to this type of seasonality (Bunis et al., 1996). Coverage on HIV/AIDS can serve to illustrate this risk, with earlier studies already observing seasonal cycles in coverage (Semetko and Goldberg, 1993; Williams and Miller, 1995), and, consequently, a decline in media attention alongside this development. This decline is also confirmed in our analysis.
The remaining seven issues included in our sample are – at very different levels of overall coverage – dominated by noise, without notable threads or storylines in coverage. Here, occasional peaks in coverage merely appear as outliers, with subsequent coverage levels quickly dropping back to pre-peak levels. These peaks in coverage can thus be interpreted as focusing events that fail to have any notable lasting impact on media, public or policy agendas. Following Neuman (1990), public attention does not correlate with media attention below a certain “threshold of public attention”. Malaria appears to be the issue that probably remains most firmly below this threshold throughout the review period; to varying degrees, however, this also appears to be the case for all other issues except climate change, poverty and HIV/AIDS. This situation calls for interventions that increase problem awareness and help creating a context in which focusing events can trigger issue-attention-cycles and corresponding policy activity.
Looking at the two environmental issues other than climate change in our sample, coverage on biodiversity and cleaner technologies are also characterized as noise rather than information. However, as indicated by pairwise correlations (Appendix 3) and co-occurrences of terms (Table 5), both issues clearly appear to be influenced by media coverage related to climate change. To some degree, this also applies to socioeconomic issues such as poverty and malaria. From a policy perspective, this interdependency between climate change and other media issues can be seen as both an opportunity and a threat for the latter. On the one hand, it is safe to assume that coverage on issues such as biodiversity and cleaner technologies has been crowded in by climate change. In other words, the dramatic increase in climate change-related media coverage may also help other environmental issues to move up the public agenda, and potentially beyond the above-mentioned threshold of public attention.
At the same time, caution should be warranted even in the case of biodiversity and cleaner technologies, as framing these challenges as part of climate change – as opposed to framing them as challenges in their own right – can also have negative consequences. Kingdon (1993: 42) provides the example of transportation of the handicapped in order to illustrate potential risks linked to the framing of a specific problem: “If transportation of the handicapped is a transportation problem, […] then low-cost and effective solutions like dial-a-ride or subsidised taxi service are appropriate. But if transportation of the handicapped is classified into the category of civil rights, then retrofitting subways for elevators and providing bus lifts is warranted. The struggle for appropriate categories makes all the difference in framing the issue.” Whilst this crowding-in process may therefore lend more visibility to an issue such as biodiversity, it might also relegate it to the status of a sub-problem associated with climate change in the perception of the general public.
A number of implications from a campaigning viewpoint follow from the diversity of patterns observed in our analysis. Here, different campaigning strategies appear suitable for different types of issues. Issues characterized by a high degree of noise and low levels of coverage (such as malaria) may predominantly require campaigns that focus on awareness raising in order to move the issue beyond the threshold of public attention. Again taking the examples of biodiversity and climate change, it is worth noting that whilst proponents of biodiversity protection have built a similar institutional infrastructure with the Intergovernmental science-policy Platform on Biodiversity and Ecosystem Services (IPBES) – which has been modelled after the IPCC with the aim to synthesize and critically evaluate knowledge related to biodiversity – this body is receiving considerably less visibility than its climate change-related equivalent. Campaigning that focuses on awareness raising therefore appears crucial in this context.
Issues dominated by noise but at higher levels of coverage might be able to concentrate on specific focusing events to trigger relevant policy discourses, for example by communicating conceivable policy options. Finally, for issues already characterized by a high degree of information, a strategy focusing on specific policy initiatives appears more suitable. For issues that reflect seasonal pattern in coverage, the analysis also provides important clues in relation to an effective timing of issue-level campaigns. For example, coverage on poverty shows seasonal cycles peaking in December; coverage on HIV/AIDS tends to peak in late November in the run-up to World AIDS Day. Therefore, the mass media can be expected to be more open to relevant poverty- and HIV/AIDS-related news during these periods, respectively. At the same time, campaigning during these peak phases may also serve to reinforce these patterns, accelerating routinization effects and ultimately leading to the general public becoming more accustomed to a given challenge and ultimately perceiving it as less urgent.
It should be kept in mind that even in the case of climate change – where a series of homogeneous peaks in coverage appear e.g. based on international policy events such as the COP meetings – a considerable diversity of patterns in coverage across and within countries can be observed across the 23 newspapers included in our sample. This clearly shows the limitations of building empirical analyses on individual media outlets, which has been a hitherto widespread practice in the literature on media coverage on climate change (for an overview see e.g. Schmidt et al., 2013). At the same time, this provides the opportunity for in-depth comparative analyses between newspapers in order to get a better understanding of the factors that determine whether or not increased levels of media attention lead to effective policy solutions.
Conclusion
In this study, based on a sample of 23 broadsheet newspapers from five different countries, we have mapped out trends and patterns in sustainability-related media coverage between 2000 and 2016. Using the agenda-setting literature and the distinction between “information” and “noise” (Black, 1986) as starting-points for our enquiry, we have developed a classification of time-series that has helped us to compare and contrast media coverage on a set of ten different sustainability challenges. We have found clearly different patterns and trajectories in coverage with regard to the two dimensions information/noise and seasonality/non-seasonality.
Our analysis makes a number of contributions to the existing literature. We have explored a wide range of sustainability challenges, thereby shedding light on the considerable diversity of sustainability-related media coverage. By employing a large sample of 23 leading newspapers, we have also uncovered clear differences between patterns identified at the level of individual newspapers and aggregate patterns across the entire sample of newspapers. As a theoretical contribution, the application of information and noise in the context of media coverage and in particular the classification of time-series enable us to better understand similarities and differences in coverage between different media issues. Crucially, we identify a divide between media coverage on socioeconomic and environmental challenges included in the analysis, and observe crowding in and crowding out processes in the interplay of climate change and other sustainability challenges.
We are aware of a number of limitations within the methodology adopted in this paper, with data availability a major factor for sample selection. All five countries included in the sample are OECD countries. Therefore, generalizations beyond these countries might not be possible. In addition, some of the key words used to track the ten sustainability issues may carry slightly different meanings when translated into different languages. However, the list of search terms was checked by experts in the field and native speakers for all three languages included in the sample. Finally, it should be noted that media coverage on any of the ten issues included in our sample will inevitably cover various different discourses linked to each of these sustainability challenges. Taking the example of climate change, overall trends may arguably not be reflected at the level of more specific climate-related challenges such as sea level rise (Yusuf et al., 2015) or climate change adaptation (Dannevig et al., 2013).
There are numerous opportunities for subsequent research. The classification of time-series developed in this study could be applied to a wide range of media analyses beyond the sustainability challenges examined in this paper. At the same time, the enquiry into sustainability-related media coverage could be extended in at least three promising ways. Future research could explore the interaction between media coverage on different challenges in more detail, most notably the questions whether and how climate change may crowd in (out) other sustainability challenges, and how this impacts the framing of these challenges. This also includes the framing of sustainability itself and the specific challenges it is commonly associated with.
In addition, the current analysis could be extended to other types of media and a more diverse set of countries, most notably with regard to non-Western contexts that have hitherto been largely neglected in the literature (see e.g. Schäfer and Schlichting, 2014; Schäfer et al., 2014). Finally, enquiries into the relationship between individual newspapers and aggregate patterns would be a promising avenue for research. This includes the role of individual newspaper characteristics and country contexts in deviating from the aggregate discourse that can be observed. For example, political alignment as well as a range of contextual factors may determine the way in which a given sustainability challenge is covered. To a certain extent, the literature focusing on media representations of climate change already offers a number of blueprints for these types of analysis which could be extended to a larger and more diverse sample.
Footnotes
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
