A tale of four cities: A semantic analysis comparing the newspaper coverage of air pollution in Hong Kong,London,Pittsburgh,and Tianjin from 2014 to 2017

Abstract

A computer-aided semantic analysis (using Linguistic Inquiry and Word Count [LIWC]) examined how newspaper coverage of air pollution from 2014 to 2017 may affect the public agenda in four cities—Hong Kong, London, Pittsburgh, and Tianjin. Results show that after controlling for the real-time air quality, the agenda-setting effect was found in Hong Kong, London, and Pittsburgh, but not Tianjin. Tianjin’s reports also contained more future-framed words but fewer present-framed words than other cities.

Keywords

agenda setting computer-aided content analysis semantic analysis newspaper coverage

Air pollution has become a global environmental calamity that has long-term adverse effects on human health. Exposure to air pollutants results in a variety of illnesses, including asthma, birth defects, bronchitis, leukemia, liver cancer, heart disease, and strokes (see De Marco et al., 2019). More than 92% of the global population was affected by air pollution and each year more than 6 million deaths worldwide were directly related to exposure to air pollutants, especially PM2.5 (i.e., inhalable particles with diameters smaller than 2.5 µm) and PM10 (i.e., inhalable particles with diameters smaller than 10 µm) (World Health Organization [WHO], 2016). Despite the identified negative outcomes of air pollution, it remains unclear how the air pollution coverage affects the public agenda, especially in a cross-cultural context.

This study uses the Linguistic Inquiry and Word Count (LIWC), a computer-aided content analysis software, to conduct semantic analyses of the newspaper reports on the air pollution in Hong Kong, London, Pittsburgh, and Tianjin from January 2014 to March 2017. The findings should provide insights into the news decision-making process in reporting environmental issues and contribute to understanding how the public agenda may be affected by the news coverage of air pollution coverage, which is, to the best of our knowledge, the first of its kind in environmental studies.

Literature Review

Four global cities’ air pollution reports were selected for this research, two of which are in China (i.e., Hong Kong and Tianjin), one in Europe (i.e., London), and one in the United States (i.e., Pittsburgh). The four cities were selected for this research because all of them are or had been heavily air-polluted. Due to the shared pollution history, residents in these cities could be sensitive to air pollution information and thus might have a strong information need regarding such environmental issues. Another reason for choosing the four cities is that Hong Kong and Tianjin are currently experiencing severe air pollution, while London and Pittsburgh have significantly decreased their environmental problems. The current air conditions in Hong Kong and Tianjin are far worse than that in London and Pittsburgh. New insights might be generated by comparing how news media report different levels of urban air pollution. Finally, the four cities have different economic, political, and media systems, ranging from relatively little government control in London and Pittsburgh, to some control like in Hong Kong, to tight control like in Tianjin. Hence, the same reality—air pollution—might be perceived differently by the news media outlets.

Air Pollution in Four Cities

Air pollution has been a major environmental concern in China, including Hong Kong and Tianjin. Hong Kong, with a population of more than 7 million, is among the most densely populated cities in the world (Hong Kong Census and Statistics Department, 2017). As a global financial center, Hong Kong’s economy heavily relies on finance and service sectors, but its air pollution has started to endanger the health of residents and drove away some foreign businesses. Levels of cancer-causing pollutants in Hong Kong peaked at 5 times higher than the acceptable level, surpassing the WHO standards for more than 15 years (Haas, 2017). Tianjin’s air pollution was even worse than that in Hong Kong. For years, Tianjin had been among China’s most polluted cities (“The 10 Polluted Cities in China,” 2016).

Comparatively, London and Pittsburgh have better air quality after they went through severe air pollution problems just decades ago. London had a long history of air pollution, which remains an environmental challenge today. Starting in the 1700s, London’s air quality had turned worse each year and reached the peak at the end of 19th century (Ritchie, 2017). In December 1952, the Great Smoke of London resulted in unprecedented morbidity and mortality, leading to more than 12,000 deaths (Bell, Davis, & Fletcher, 2004). The environmental crisis directly led to the Clean Air Act in 1956 that restricted coal-burning and gradually replaced coal with alternative heating sources (Klein, 2012).

In the United States, Pittsburgh had been a heavily air-polluted city for over a century from the 1870s to the 1980s, which was known as the “Smoky City” or “Hell with the Lid Off” (“Pittsburgh, the ‘Smoky City,” 2015). Since its war against air pollution started in the 1940s, Pittsburgh’s industries and residents began to adopt smokeless fuels, such as anthracite coal and natural gas, resulting in better air quality (Davidson, 1979). Significant industrial transformation happened during the 1980s when commercial, retail, and residential districts gradually replaced industrial sites (“Pittsburgh, the ‘Smoky City,” 2015). Other U.S. cities also have decreased their air pollution, however, and Pittsburgh still routinely ranks among the 10 worst U.S. cities for air pollution (and often the only one outside of California). But today, a variety of industries, such as health care, technology, and financial services, have been developed and the steel industry became the city’s history.

Although the air quality in London and Pittsburgh is better than decades ago, their history of air pollution has been regularly mentioned and reported in the news. Thus, the current air quality conditions may shape how air pollution is reported in each of the four cities and the news coverage’s agenda-setting effects may vary due to the different levels of air pollution.

Agenda-Setting Theory

The agenda-setting theory should be an appropriate theoretical framework for this study. The theory assumes a significant correlation between the news media agenda and the public agenda as the importance attributed to certain issues by news media would considerably influence the salience placed on these issues by mass audiences (McCombs & Shaw, 1972). Since the publication of the classic Chapel Hill study (McCombs & Shaw, 1972), the agenda-setting effect has been consistently confirmed in hundreds of empirical studies conducted for numerous media types and public issues in a variety of contextual settings (McCombs & Reynolds, 2009). The agenda-setting effect was also found in the field of environmental communication. Prior studies had discovered a significant relationship between news media’s focus on the environmental issues and the salience of those issues in the public agenda (Shanahan, Morgan, & Stenbjerre, 1997).

For this study, the four cities’ air quality data were collected from local air quality monitoring stations, which provide the real-time air quality data to the public through the Internet. Real-time conditions, defined as “the actual prominence of the specific issue in reality, the indicators of current conditions, or the measure of the importance of an issue in reality, usually based on statistics” (Ader, 1995, p. 300), are important factors to consider when evaluating environmental news effects. In his study of the environmental pollution coverage in The New York Times from 1970 to 1990, Ader (1995) found a negative correlation between the news media agenda and the real-time pollution conditions, suggesting that when the air quality improved, news media covered more about air pollution. In this longitudinal study, Ader (1995) argued that the environmental pollution was an unobtrusive issue that would manifest a significant agenda-setting effect. In the cases of Hong Kong and Tianjin, the air pollution was so obtrusive that it interrupted local people’s daily lives and endangered their health. In this sense, it is important to examine if the real-time air quality data could be a confounding variable that influences the public agenda through the coverage of air pollution. To researchers, an objective assessment of the reality is difficult (McLeod, Becker, & Byrnes, 1974), which may explain why few prior studies included real-time conditions into the content analysis research. Hence, the first research question was proposed:

RQ1:Controlling for the real-time air quality conditions, what is the relationship between the news media agenda involving air pollution and the public agenda in the four cities?

Second-Level Agenda-Setting

News media may not only tell us what to think about, but also how to think. The transfer of the attribute salience from news media to the public is known as second-level agenda-setting effects (McCombs & Shaw, 1993). When the traditional agenda-setting centers on the “issue” salience, the second-level agenda-setting focuses more on the “attribute” salience of an issue (Kim & McCombs, 2007). On one hand, news media have the power to direct our attention to an object, while, on the other hand, they can guide our attention to certain characteristics and properties of the object, which are known as “attributes” (Kim & McCombs, 2007).

This research thus explores the impact of real-time air quality conditions on the tone of air pollution reports, that is, the positivity or negativity of the reports. Positive and negative evaluations are the fundamental aspects of any human language (Osgood, 1959). Both air pollution index (API) and temperature were found to have a significantly positive relationship with the negativity of the U.S. journalists’ coverage of the Beijing Olympics (Zhong & Zhou, 2012). Findings suggested that journalists’ decision-making was influenced by factors that were previously ignored, such as the environmental cues like weather conditions (Zhong & Zhou, 2012). The rationale was that weather conditions might impact journalists’ mood, which in turn would alter their positive and negative word choices (Zhong & Zhou, 2012). However, the study rested only on a particular context, that is, U.S. journalists’ sports reporting in a foreign city (Beijing), and no study was found to investigate such an association in other settings. Given the different air pollution conditions in the four cities, this study explores the possible differences of emotional words used in reporting the four cities’ air pollution. Hence, two more research questions were formulated:

RQ2:What is the relationship between real-time air pollution conditions and the tone of the air pollution coverage?

RQ3:How do the four cities differ in terms of the emotional words used in the air pollution reports?

Chinese news media often linked the issue of air pollution to the country’s economic development. A 2013 content analysis of air pollution reporting showed that 35% of articles in China Daily linked air pollution to societal and economic factors, whereas only 17% of articles mentioned its impact on human health, and only one article linked it to the ecosystem (Zu, 2015). We were unable to find any study that analyzed if air pollution reporting is correlated with the economic reporting, so this study thus poses the following research question:

RQ4:Do the tones of news reports on air pollution and economic issues in the four cities share similarities in terms of emotional words used?

When covering environmental concerns, journalists tend to view time orientation as another important attribute. The time orientation is important as it discloses the news media’s implicit attention to certain dimensions of the news issues. Chyi and McCombs’s (2004) content analysis of the Columbine school shooting reports showed that the vast majority of stories used a present frame. Following the same rationale, Houston, Pfefferbaum, and Rosenholtz (2012) examined the disaster coverage in U.S. newspapers and network evening news broadcasts, whose results confirmed that U.S. news media mainly focused on the present in reporting disaster events. Many such studies have been about the U.S. or other Western journalistic practices, with little effort devoted to analyzing the time orientation in other countries’ news reporting. Hence, the final research question was proposed to explore the time orientation in the four cities’ air pollution coverage:

RQ5:Is the four cities’ air pollution coverage different from each other in terms of the time orientation—the reports mainly using a past, present, or future frame?

Method

Four air pollution reporting corpora were collected for this research, namely Hong Kong’s corpus collected from the South China Morning Post, London’s corpus from The Times, Pittsburgh’s corpus from the Pittsburgh Post-Gazette, and Tianjin’s corpus from Beijing-based China Daily, which has a Tianjin bureau (but not a Tianjin edition). LexisNexis analyzed and assigned index terms that helped identify the attributes of news articles in its database, including industry, subject, geography, and people being related to and mentioned in the articles (LexisNexis, 2017). All articles were then downloaded from the LexisNexis database after they were identified by a subject index of “air pollution.”

Several advantages emerged when the sample was collected by using the subject index of “air pollution,” rather than relying on keyword searching. First, journalists often used interchangeable words referring to air pollution in the news reports, such as air quality, air pollutant, haze, and emission. Some news articles that did not contain the keyword “air pollution” were included in the sample as their themes were about air pollution. This was particularly evident in the London and Pittsburgh corpora, where the current air pollution situation was no longer as bad as before, and as a result journalists reported more on other aspects of air pollution. For example, one Times story reported on pollution caused by diesel car taxes was included in the sample based on the subject index approach, though none of the keywords about air pollution was used. Similarly, a Pittsburgh Post-Gazette article introduced the “Particle Falls”—an installation that can stream colorful light that looked like rainfall burst into flames when pollution particles reached a certain level—to draw residents’ attention to possible air pollution. The article was also included in the sample as it reported local air pollution. In other scenarios, a report mentioning the keyword “air pollution” was excluded as the article centered on other urban issues. Therefore, data gathered through the subject index approach could be more relevant and accurate than keyword searching.

Finally, a location index was also used in combination with the subject index to search the reports on each city’s air pollution (e.g., “air pollution” + “Tianjin” in searching for Tianjin’s air pollution reports). All news articles that met the search criteria were included in the sample, which were those published from January 2014 to March 2017. This time period was chosen because, in LexisNexis, the earliest China Daily reports on Tianjin’s air pollution were available starting in January 2014 and ended in March 2017. The 39-month period thus became the cut-off phrase for all four corpora under study, which consists of 1,982 newspaper reports.

The air pollution salience in a given newspaper (i.e., the news media agenda) was operationalized as the average keywords in each report in each month. Specifically, the LIWC counted the number of air pollution-related keywords, that is, air pollution, air pollutant, smog, and so on, and divided it by the total number of the words used in each article on air pollution. The quotient in each article was added up and the sum was then divided by the total number of articles published by the respective newspaper in each month to get the average number of keywords in a given month for that newspaper. Therefore, the higher the number was, the more salient the air pollution issue was in that month in the city. A separate set of corpora were collected that reported the four cities’ economic issues by searching the index term “economy & economic indicators.” These articles were also downloaded month by month for the same period (January 2014 to March 2017). A comparison could thus be made between the reports on air pollution and economic activities in each city.

The LIWC Analysis

All corpora were processed by the LIWC software (Pennebaker, Boyd, Jordan, & Blackburn, 2015). Compared to manual content analysis procedures featured by a series of time-consuming and tedious human-coded methods, computer-aided content analysis offered several distinct advantages, such as the stability and comparability of coding rules leading to more accuracy in the research findings, perfect coder reliability in applying the rules to texts, and exceptional speed in processing huge quantities of text data (Zhong & Zhou, 2012). The LIWC user can input any text files into the software and it then counts and calculates the percentage of words based on a variety of pre-programmed linguistically and psychologically meaningful categories. Currently, LIWC has 93 word categories, such as positive emotions, negative emotions, and future-framed words (Tausczik & Pennebaker, 2010). The validity and reliability of LIWC calculations have been well demonstrated in a series of studies (Tausczik & Pennebaker, 2010).

For our research, the percentage of words in each category was computed by the LIWC with the unit of analysis being all air pollution–related articles published in 1 month. For instance, China Daily had 24 articles that covered Tianjin’s air pollution issues in March 2017, and the LIWC calculated the percentage of each word categories accounted for the total number of words in the articles. A higher number in one category would indicate that articles in that month were more prone to this psychological state that this word category referred to. Thus, a report containing a higher number of positive words was regarded as more prone to a positive psychological state. Finally, the LIWC computed the six key variables by the following formulas:

\begin{array}{l} Affective word frequencies = affect words / total words; \\ Positive = Positive Words / Total Words; \\ Negative = Negative Words / Total words; \\ Past = Past - Framed Words / Total Words; \\ Present = Present - Framed Words / Total Words; \\ Future = Future - Focused words / Total Words; \end{array}

Based on the above variables, four more variables were computed in the LIWC:

\begin{array}{l} Positivity = Positive - Negative; \\ Past - relativity = Past / (Past + Present + Future); \\ Present - relativity = Present / (Past + Present + Future); \\ Future - relativity = Future / (Past + Present + Future) . \end{array}

To avoid semantic biases in processing the words in the news reports, some words were particularly processed by LIWC. For example, all the words “May” were replaced by the word “Mei” in the articles published in the May months before being processed. The procedure would prevent the LIWC from mistakenly categorizing the word “May” as a future-focused word, resulting in a disproportionately high percentage of future-framed words in each of the May month samples.

The Web Search Index

The ubiquity of Internet search among the public has made the search records an accurate, real-time, and easily accessible indicator of the public attention to specific issues (Desjardins, 2018; DMR, 2019). Researchers used to measure the salience of a public issue by conducting polls, monitoring public opinions, or analyzing public attention and concerns (Kwak, An, Salminen, Jung, & Jansen, 2018). Now they are increasingly assessing the public agenda by mining Internet search data, which has been repeatedly tested and validated in numerous studies. Scharkow and Vogelgesang (2011), for instance, used data from Google Insights for Search to measure the public agenda and concluded that the approach was a promising tool for agenda-setting research. Mellon (2014) compared Google Trends data with the Gallop’s “Most Important Problem” poll data from 2004 to 2010 and found that Internet search data were a reliable way to measure the issue salience as part of the public agenda. Following them, the public agenda of air pollution in the four cities was measured by the Internet search data.

The keywords searched by the Internet users in the four cities from January 2014 to March 2017 were used as the indicator of the public agenda. For Hong Kong, London, and Pittsburgh, the English keyword “air pollution” was used and for Tianjin the Chinese keyword of 空气污染 (air pollution) was used. The search results of “air pollution” were collected from Google Trends (https://trends.google.com/trends), while the search results of “空气污染” came from a Chinese search engine, Baidu Index (www.index.baidu.com). A monthly average search index was thus created for each city’s air pollution issues, reflecting how frequently local residents in the four cities searched the keywords “air pollution” in English or Chinese.

Air Quality Indicators

Three measures of air quality data were collected for each city, namely the monthly average PM2.5, PM10, and Air Quality Index (AQI), which are the primary sources for health concerns (U.S. Environmental Protection Agency [EPA], 2016). While Mainland China and the United States employed the same AQI standards (some studies suggested differences exist between the two countries; see Andrews, 2014), Hong Kong and the United Kingdom used different criteria for the AQI. Hong Kong used the Air Quality Health Index ranging from 1 to 10+ and London employed the Daily Air Quality Index ranging from 1 to 10. The AQIs from the four cities were not comparable directly, but they provided useful information on the air conditions in these cities. In addition to the AQI, this study uses two more air quality indicators—PM2.5 and PM10—for measuring air pollution. The three indicators are significantly correlated, thus providing more comprehensive information on the air quality. In this study, the AQI, PM2.5, and PM10 data came from four sources: the Hong Kong data were from the websites of Hong Kong Environmental Protection Department: https://cd.epic.epd.gov.hk/EPICDI/air/station/?lang=en, the London data from the website: https://data.london.gov.uk/dataset/london-average-air-quality-levels, the Pittsburgh data from the U.S. Environmental Protection Agency: https://aqs.epa.gov/aqsweb/airdata/download_files.html#AQI, and the Tianjin data from the website aqistudy.cn .

Results

Table 1 reports all the means and standard deviations of PM2.5, PM10, AQI, news media agenda, public agenda, and the percentage of emotional words in the sample.

Table 1

Descriptive Statistics of Air Quality Indicators, Media Salience Score, Public Salience Score, and the Percentage of Emotional Words for Four Cities, by Month, Mean (SD)

	Hong Kong	London	Pittsburgh	Tianjin
PM2.5	25.64 (10.91)	15.00 (4.24)	9.79 (1.88)	75.92 (25.09)
PM10	39.21 (15.49)	24.51 (4.59)	19.87 (5.11)	121.39 (34.58)
AQI	3.46 (0.67)	3.26 (0.27)	44.95 (7.39)	110.23 (26.99)
Media agenda	1.69 (1.11)	1.42 (1.10)	0.55 (0.48)	4.04 (2.47)
Public agenda	23.16 (7.88)	2.80 (0.40)	29.90 (11.30)	96.44 (24.91)
Affect	3.36 (0.40)	3.44 (0.62)	3.16 (0.51)	2.80 (0.40)
Positive	1.89 (0.28)	1.69 (0.45)	2.03 (0.45)	1.58 (0.44)
Negative	1.41 (0.34)	1.72 (0.48)	1.09 (0.42)	1.16 (0.29)
Positivity	0.47 (0.47)	−0.03 (0.71)	0.93 (0.72)	0.42 (0.62)

Note. All sample sizes N = 39. AQI = Air Quality Index, Media agenda = number of keywords/number of articles in 1 month, Public agenda = public web search index in a month.

Agenda-Setting and Real-Time Conditions

The Pearson correlation coefficient analyses showed that both the news media and public agendas of air pollution issues were significantly correlated in each of the four cities, for Hong Kong, r₍₃₉₎ = .52, p < .001, for London r₍₃₉₎ = .56, p < .001, for Pittsburgh, r₍₃₉₎ = .51, p <.001, and for Tianjin, r₍₃₉₎ = .51, p < .001. The results indicate that the news media agenda had a significant impact on the public agenda in the four cities. To answer RQ1, more correlation analyses were conducted after controlling for the real-time air conditions, that is, PM2.5 and PM10. The results showed that for Tianjin, the media agenda was no longer correlated with the public agenda of air pollution, r₍₃₆₎ = .17, p = .305, after controlling for PM2.5. However, this relationship was still significant for Hong Kong, r₍₃₆₎ = .40, p < .05, London, r₍₃₆₎ = .46, p < .01, and Pittsburgh, r₍₃₆₎ = .47, p < .01.

After controlling for the real-time PM10 data, the analysis showed a similar trend: for Tianjin, the news media’s agenda-setting effects no longer existed, r₍₃₆₎ = .26, p = .119. But the effects remained significant for Hong Kong, r₍₃₆₎ = .41, p < .01, London, r₍₃₆₎ = .46, p < .01, and Pittsburgh, r₍₃₆₎ = .43, p < .01. Since the AQI from the four cities were measured in different scales, no analysis could be conducted based on the absolute values of AQI. Considering the overall significant correlation between air quality indicators of PM2.5, PM10, and AQI in the four cities, it can be inferred that the same results would show a similar trend after controlling for the AQI.

Air Quality and Affective Tone

Again, RQ2 asks whether the tone of air pollution reporting (i.e., positivity or negativity) would be related to the air quality conditions. The Pearson correlation coefficient analyses were conducted to test the relationships between three air quality indicators and the usage of emotional words, positive affect words (e.g., improving, successfully, greater), negative affect words (e.g., fight, serious, low), and the positivity score in the air pollution articles (see Table 2). The three air quality indicators in Tianjin were negatively correlated with both the positive word frequency and the positivity level in the air pollution reporting. However, no significant relationship was found for Hong Kong, London, or Pittsburgh. Only one significantly positive relationship emerged: between the affect word frequency and Pittsburgh’s AQI, r₍₃₉₎ = .46, p < .01. The results show that when Tianjin’s air pollution worsened, more news articles were produced to cover it and the tone of news coverage turned more negative, or vice versa. In contrast, the trend was not found in the other newspapers.

Table 2

Correlations Between Air Quality Indicators and Affective Word Frequencies

	Hong Kong			London			Pittsburgh			Tianjin
	PM2.5	PM10	AQI	PM2.5	PM10	AQI	PM2.5	PM10	AQI	PM2.5	PM10	AQI
Affective word frequencies	.05	.04	−.10	.05	.11	.20	.26	.18	.46**	−.21	−.28	−.21
Positive	−.05	−.05	−.08	−.00	.01	−.09	.28	.14	.28	−.39*	−.43**	−.38*
Negative	.10	.08	−.05	.08	.14	.17	.02	.08	.26	.30	.91**	.28
Positivity	−.10	−.08	−.01	−.06	−.09	−.06	.17	.05	.02	−.42**	−.42**	−.40*

Note. AQI = Air Quality Index. *p < .05, **p < .01, ***p < .001, df = 39.

Emotional Words in the Air Pollution Reports

To answer RQ3, analyses of variance (ANOVAs) were conducted to compare positivity scores of the air pollution coverage in the four cities with the city as the independent categorical variable and positivity scores as the dependent continuous variable. The findings show that the positivity scores of the newspaper coverage were significantly different between the four cities, F (3, 152) = 14.98, p < .001, $η_{p}^{2} = . 23$ . The Bonferroni post hoc tests showed that the tone in the Pittsburgh air pollution reports was the most positive, M = 0.93, SD = 0.72, and it was significantly higher than Hong Kong, M = 0.48, SD = 0.48, p < .01, London, M = −0.03, SD = 0.71, p < .001, or Tianjin, M = 0.42, SD = 0.62, p < .001. The tone in the London air pollution reports was the most negative, which was significantly lower than Hong Kong, M = 0.48, SD = 0.48, p < .001, Pittsburgh, M = 0.93, SD = 0.72, p < .001, or Tianjin, M = 0.42, SD = 0.62, p < .01. No tone difference was detected between the air pollution reports in Tianjin and Hong Kong, which might be due to the similar air pollution conditions in the two cities.

Air Pollution Reporting and Economic Reporting

Again, RQ4 asks about any relationships between the four cities’ air pollution and economic reports in terms of affective words used. As shown in Table 3, bivariate correlation analyses reveal that the affective word frequencies in the air pollution reports and economic stories were positively correlated in Hong Kong and Tianjin. This means that the emotional words used in the air pollution reporting were highly correlated to those in the economic reporting for Hong Kong and Tianjin. In contrast, the same trend did not exist for the two types of news reports on London or Pittsburgh.

Table 3

Correlations of Affective Word Frequencies Between Air Pollution Reporting and Economy Reporting

	Hong Kong	London	Pittsburgh	Tianjin
Affective word frequencies	.32*	.03	.10	.48***
Positive	.02	.03	.03	.54***
Negative	.45**	−.10	.19	.39*
Positivity	.22	−.04	.09	.50***

Note. *p < .05, **p < .01, ***p < .001, df = 39.

Time-Orientation Comparison

The time-orientation variable reflects the amount of news media attention devoted to one type of time-orientation words compared with the other types of time-orientation words regardless of any systematic bias. To answer RQ5, the percentages of past-framed (e.g., were, been, said), present-framed (e.g., work, gets, be), and future-framed word frequencies (e.g., will, plan, prepare) in the four corpora of air pollution reports were compared through a series of ANOVA analyses. The variables of “past relativity,” “present relativity,” and “future relativity” were also compared in the analyses. Results showed that Tianjin’s air pollution reports (M = 3.94, SD = 0.50) had significantly less present-framed words than that in the Hong Kong sample (M = 5.14, SD = 0.92, p < .001), the Pittsburgh sample (M = 5.01, SD = 1.00, p < .001), or the London sample (M = 5.52, SD = 0.96, p < .001), F (3, 152) = 23.72, p < .001, $η_{p}^{2} = . 32$ . Meanwhile, the Tianjin reports (M = 1.28, SD = 0.06) contained significantly more future-framed words than those in the Hong Kong sample (M = 0.94, SD = 0.06, p < .001), F (3, 152) = 3.42, p < .05, $η_{p}^{2} = . 06$ . But no significant relationship existed between the Tianjin sample (M = 2.81, SD = 0.34) and other three samples in terms of using past-framed words, for Hong Kong (M = 3.12, SD = 0.57), for Pittsburgh (M = 2.95, SD = 0.90), and for London (M = 3.03, SD = 0.75), F (3, 152) = 1.55, p = .21, $η_{p}^{2} = . 03$ .

The results also reveal that the Tianjin sample (M = 49.54, SD = 4.46) also had significantly lower present relativity score than the Hong Kong sample (M = 55.39, SD = 6.72), Pittsburgh (M = 55.05, SD = 8.60) or the London sample (M = 57.36, SD = 7.09), F (3, 152) = 9.29, p < .001, $η_{p}^{2} = . 16$ . It also had a significantly higher future relativity score (M = 14.96, SD = 2.80) than Hong Kong (M = 10.49, SD = 2.40), Pittsburgh (M = 12.31, SD = 4.99), and London (M = 10.93, SD = 3.53), F (3, 152) = 12.43, p < .001, $η_{p}^{2} = . 20$ . Yet the Tianjin sample had the highest past-framed score, though the difference between four samples was not significant, F(3, 152) = 1.89, p = .13, $η_{p}^{2} = . 04$ .

Discussion

A number of points from the data require some comment. First, the time-orientation analysis showed that Tianjin’s air pollution reports focused more on the future and less on the present than other cities under study. These results imply that the Tianjin reports tried to paint a rosier picture of air quality in the future so to alleviate local residents’ concerns on the imminent air pollution at hand.

Second, again, after controlling the air quality condition, the agenda-setting effect disappeared for Tianjin. The previous significant relationship between media agenda and public agenda could be largely accounted for by the real-time air quality conditions in Tianjin. This indicated that the impact of Tianjin’s media agenda on the public opinion was not as strong it seemed, which may be partially due to the tight control of Chinese news media or, more likely, due to China Daily being a national, English-language newspaper that probably is not well read by Tianjin’s Chinese residents.

Third, that non-significant agenda-setting effect in Tianjin certainly deserves more research, but the result may reflect concerns about trustworthiness of Chinese news media outside of Hong Kong. When public opinion was less affected by news media than by personal experiences, the credibility of Chinese news media among the public could be assumed to be low, which Tsfati (2003) called “media skepticism.” Affective word frequencies results in Tianjin further strengthened the argument that real-time air quality influenced both the news media agenda and public agenda in Tianjin.

Fourth, that air pollution reporting only in Tianjin had significantly more future-framed words and fewer present-framed words than the other three corpora requires analysis. It may result from differences in media systems between Tianjin and other cities (see “Limitations and Future Studies” section).

Previous studies (e.g., Chyi & McCombs, 2004; Houston et al., 2012) found that at least U.S. news media focus more on the present. China Daily’s reporting on Tianjin’s air pollution focused significantly more on the future and less on the present than the other three newspapers. This indicated that China Daily tried to avoid covering the Tianjin air pollution in detail, but focused more on the government’s determination to search for future solutions.

Limitations and Future Studies

Several limitations should be noted when interpreting the findings. First, all relationships found in this study may not be interpreted as causal relationships. The salience of air pollution in the news media was correlated with the public salience in the same month without including any time lag. This was mainly because the time lag for an issue like air pollution would likely be as short as a couple of days. Thus, the unit of analysis, ideally, should be weekly or daily air pollution data, not the monthly data used in this study. Further analysis may explore the question of causality by using weekly or daily data.

Second, the air pollution coverage period from January 2014 to March 2017 was selected based on the availability of archives in LexisNexis, in which the reports on Tianjin were archived starting in January 2014. Future studies could include different time periods or even compare the historic data to more recent data. For example, a study could compare news reports in the 1950s when Pittsburgh air pollution was severe to those in recent years from Hong Kong and Tianjin. Such a study should generate more in-depth findings and thus make a significant contribution to the understanding of news media portrayal of air pollution. Finally, this study identifies and explores three variables that could potentially moderate the agenda-setting effect: the real-time air pollution condition, the affective tone, and the media trustworthiness. Future research may also examine how and in what contexts these variables may alter agenda-setting effects.

Third, the newspapers were not directly comparable: the Pittsburgh Post-Gazette is local while the others are national, even international, in readership and influence; China Daily is not based in Tianjin while the other newspapers are based in the cities studied; and neither China Daily nor South China Morning Post are published in Chinese (and nearly 50% of Hong Kong residents speak English compared with less than 1% of China’s residents outside of Hong Kong).

Conclusion

An important theoretical contribution of this comparative study is shedding light on issue obtrusiveness in agenda-setting theory. In Ader’s (1995) study of agenda-setting, air pollution was regarded as an unobtrusive issue, and the correlation between the news media agenda and public agenda was strengthened when the effect of reality was controlled. Contrary to that result, this study chose four cities where air pollution was or had been an obtrusive issue. The result indicated that the correlation between the news media salience and public salience of air pollution was weakened when real-time air quality was controlled. This was particularly noticeable in Tianjin, where air pollution was the worst and most obtrusive, and the agenda-setting effect disappeared as long as the real-time air quality was controlled. Thus, real-time air quality was found to affect both the news media and public agendas in Tianjin.

The analyses discovered the scale of the negative tone of air pollution reports from high to low was in the order: London, Tianjin, Hong Kong, and Pittsburgh. Interestingly, the ranking of the strength of agenda-setting correlations for the four cities from strongest to the weakest followed the same order. This suggested that affective attributes had agenda-setting effects, which is consistent with prior studies (e.g., Wu & Coleman, 2009). Specifically, when news media covered air pollution more negatively, the public would pay more attention to the issue as it was made more salient in the news media agenda.

This study also identifies the relationship between the air pollution reporting and economic reporting. Emotional words used in the air pollution and economic reports were significantly correlated for both the Tianjin and Hong Kong samples, but not for the London and Pittsburgh samples. This might be due to the fact that many news articles in Tianjin and Hong Kong corpora discussed air pollution in the context of the economy, which seldom happened in the London and Pittsburgh corpora. The close relationship between air pollution and the economy can be interpreted as framing the air pollution as a problem resulting from humans’ economic activities rather than natural factors. In this study, both China Daily and South China Morning Post linked the two issues closely, but not so much in The Times or Pittsburgh Post-Gazette.

Generally, the professional and academic fields of environmental communication require a better understanding of news media portrayal of, and public attitudes toward, air quality because both public involvement and news media perceptions are crucial to successfully curbing air pollution. As long as urban air pollution continues to be a major global risk, more comparative studies like this are needed to systematically investigate how news media and the public perceive air pollution risks. Then, the results of such studies need to get into the hands of policymakers and the news media. These studies’ findings should empower residents to play a more significant role in local pollution-control strategies, which are also relevant to policymakers in regulating air pollution. This is never an easy task, and this study represents such an effort in achieving it.

Footnotes

ORCID iD

Tongxin Sun

References

Ader

C. R.

(1995). A longitudinal study of agenda setting for the issue of environmental pollution. Journalism & Mass Communication Quarterly, 72, 300-311. doi:10.1177/107769909507200204

Andrews

S. Q.

(2014, March 27). China’s air pollution reporting is misleading. Retrieved from https://www.chinadialogue.net/article/show/single/en/6856-China-s-air-pollution-reporting-is-misleading

Bell

M. L.

Davis

D. L.

Fletcher

(2004). A retrospective assessment of mortality from the London smog episode of 1952: The role of influenza and pollution. Environmental Health Perspectives, 112(1), 6-8. doi:10.1289/ehp.6539

Chyi

H. I.

McCombs

(2004). Media salience and the process of framing: Coverage of the Columbine school shootings. Journalism & Mass Communication Quarterly, 81(1), 22-35. doi:10.1177/107769900408100103

Davidson

C. I.

(1979). Air pollution in Pittsburgh: A historical perspective. Journal of the Air Pollution Control Association, 29(10), 1035-1041. doi: 10.1080/00022470.1979.10470892

De Marco

Proietti

Anav

Ciancarella

D’Elia

Fares

. . . Leonardi

(2019). Impacts of air pollution on human and ecosystem health, and implications for the National Emission Ceilings Directive: Insights from Italy. Environment International, 125, 320-333. doi:10.1016/j.envint.2019.01.064

Desjardins

(2018, April 20). This chart reveals Google’s true dominance over the web. Retrieved from https://www.visualcapitalist.com/this-chart-reveals-googles-true-dominance-over-the-web/

DMR. (2019). 90 amazing Baidu statistics and facts (2019): By the numbers. Retrieved from https://expandedramblings.com/index.php/baidu-stats/

Haas

(2017, February 16). Where the wind blows: How China’s dirty air becomes Hong Kong’s problem. The Guardian. Retrieved from https://www.theguardian.com/cities/2017/feb/16/hong-kong-death-trap-dirty-air-pollution-china

10.

Hong Kong Census and Statistics Department. (2017, August 17). Population. Retrieved from https://www.censtatd.gov.hk/hkstat/sub/so20.jsp

11.

Houston

J. B.

Pfefferbaum

Rosenholtz

C. E.

(2012). Disaster news: Framing and frame changing in coverage of major U.S. natural disasters, 2000–2010. Journalism & Mass Communication Quarterly, 89(4), 606-623. doi:10.1177/1077699012456022

12.

Kim

McCombs

(2007). News story descriptions and the public’s opinions of political candidates. Journalism & Mass Communication Quarterly, 84(2), 299-314. doi:10.1177/107769900708400207

13.

Klein

(2012, December 6). The great smog of 1952. History. Retrieved from http://www.history.com/news/the-killer-fog-that-blanketed-london-60-years-ago

14.

Kwak

Salminen

Jung

S. G.

Jansen

B. J.

(2018). What we read, what we search: Media attention and public attention among 193 countries. WWW2018. Retrieved from https://arxiv.org/pdf/1802.06437.pdf

15.

LexisNexis. (2017). LexisNexis® Publisher user guide. Retrieved from http://help.lexisnexis.com/tabula-rasa/publisher/userguide-ditamap-pdf?lbu=US&locale=en_US&audience=intranet

16.

McCombs

M. E.

Reynolds

(2009). How the news shapes our civic agenda. In Oliver

M. B.

(Ed.), Media effects: Advances in theory and research (pp. 1-16). New York, NY: Taylor & Francis.

17.

McCombs

M. E.

Shaw

D. L.

(1972). The agenda-setting function of mass media. Public Opinion Quarterly, 36, 176-187. doi:10.1086/267990

18.

McCombs

M. E.

Shaw

D. L.

(1993). The evolution of agenda-setting research: Twenty-five years in the marketplace of ideas. Journal of Communication, 43(2), 58-67.

19.

McLeod

J. M.

Becker

L. B.

Byrnes

J. E.

(1974). Another look at the agenda-setting function of the press. Communication Research, 1, 131-166. doi:10.1177/009365027400100201

20.

Mellon

(2014). Internet search data and issue salience: The properties of Google Trends as a measure of issue salience. Journal of Elections, Public Opinion and Parties, 24(1), 45-72. doi:10.1080/17457289.2013.846346

21.

Osgood

C. E.

(1959). The representational model and relevant research methods. In Pool

I. d. S.

(Ed.), Trends in content analysis (pp. 33-88). Urbana: University of Illinois Press.

22.

Pennebaker

J. W.

Boyd

R. L.

Jordan

Blackburn

(2015). The development and psychometric properties of LIWC2015. Austin: University of Texas at Austin.

23.

Pittsburgh, the “Smoky City.” (2015). Pittsburgh’s Dark History. Retrieved from https://popularpittsburgh.com/darkhistory/

24.

Ritchie

(2017, June 20). What the history of London’s air pollution can tell us about the future of today’s growing megacities? Retrieved from https://ourworldindata.org/london-air-pollution/

25.

Scharkow

Vogelgesang

(2011). Measuring the public agenda using search engine queries. International Journal of Public Opinion Research, 23, 104-113. doi:10.1093/ijpor/edq048

26.

Shanahan

Morgan

Stenbjerre

(1997). Green or brown? Television and the cultivation of environmental concern. Journal of Broadcasting & Electronic Media, 41, 305-323. doi:10.1080/08838159709364410

27.

Tausczik

Y. R.

Pennebaker

J. W.

(2010). The psychological meaning of words: LIWC and computerized text analysis methods. Journal of Language and Social Psychology, 29, 24-54. doi:10.1177/0261927X09351676

28.

The 10 most polluted cities in China. (2016, September 7). Retrieved from https://china-underground.com/2016/09/07/10-polluted-cities-china/

29.

Tsfati

(2003). Does audience skepticism of the media matter in agenda setting? Journal of Broadcasting & Electronic Media, 47, 157-176. doi:10.1207/s15506878jobem4702_1

30.

U.S. Environmental Protection Agency. (2016). Particulate matter (PM) basics. Retrieved from https://www.epa.gov/pm-pollution/particulate-matter-pm-basics#PM

31.

World Health Organization. (2016). WHO releases country estimates on air pollution exposure and health impact. Retrieved from http://www.who.int/mediacentre/news/releases/2016/air-pollution-estimates/en/

32.

H. D.

Coleman

(2009). Advancing agenda-setting theory: The comparative strength and new contingent conditions of the two levels of agenda-setting effects. Journalism & Mass Communication Quarterly, 86, 775-789. doi:10.1177/107769900908600404

33.

Zhong

Zhou

(2012). “Under the weather”: The weather effects on U.S. newspaper coverage of the 2008 Beijing Olympics. Mass Communication and Society, 15, 559-577. doi:10.1080/15205436.2012.677091

34.

(2015). Framing pollution: An analysis of the coverage of air pollution in China in 2013 by ChinaDaily.com (Master’s thesis, Ohio University, Athens). Retrieved from https://etd.ohiolink.edu/!etd.send_file?accession=ohiou1429698709&disposition=inline