Estimating the Weights of Media Tonalities in the Measurement of Media Coverage of Corporations

Abstract

This study estimates the weights of media tonalities in the measurement of media coverage of corporations by using linear regression analysis. Two new measures—Three Factor Media Favorability (TFMF) index and Four Factor Media Favorability (FFMF) index—are developed based on these estimations. These two measures are compared with other linear function measures of media coverage of corporations. Two types of measures—the measure of media favorability, and the composite measure of media favorability and visibility—are differentiated based on the different roles of media visibility in the formulas. Empirical analyses find that TFMF index and FFMF index have a certain degree of relative advantage over other measures of media favorability in predicting corporate reputation, and a composite measure of media favorability and visibility has a certain degree of relative advantage over the measures of media favorability in predicting corporate reputation. The analyses are based on a content analysis of 2,817 news articles from both the elite newspapers and local newspapers.

Keywords

measurement weight media tonality agenda-setting corporate reputation

The media coverage of corporations has increased rapidly since the 1980s (Henriques, 2000; Jones, 1987; Tunstall, 1996). The expansion of corporate news has not suffered from the 2008 financial crisis that affected millions of Americans. In contrast to the rapid decline of other major newspapers, the flagship business newspaper—The Wall Street Journal—has maintained a steady growth of circulation and become the largest newspaper in the U.S. by circulation (The Alliance for Audited Media, 2013). The financial crisis raised the demand of business news as people tend to seek more information in order to reduce the uncertainties in economic activities (Davis, 2000).

The role of news in the business world has attracted scholars for a long time. In particular, scholars in the management discipline have investigated the influence of media coverage on corporate reputation since the 1990s (e.g., Deephouse, 2000; Fombrun & Shanley, 1990; Wartick, 1992, 2002). Following these attempts, scholars in the communication discipline have been engaged in the same research study by applying the agenda-setting theory (e.g., Carroll & McCombs 2003; Kiousis, Popescu, & Mitrook, 2007; Meijer, 2004). Nevertheless, the empirical findings of the relationship between media coverage and corporate reputation are mixed and far from conclusive.

These inconsistent findings might be caused by multiple reasons, such as news data sources, industry sectors, the measurement of corporate reputation, and the measurement of media coverage. The current study is focused on the measurement of media coverage. In particular, it examines whether the measures with the weights estimated by regression analysis are superior to the measures with the subjectively assigned weights. The criterion of this examination is predictive validity, which is the degree to which a measure is correlated to the future criterion/observable facts (Carmines & Zeller, 1979; Krippendorff, 1980). Two new measures—Three Factor Media Favorability (TFMF) index and Four Factor Media Favorability (FFMF) index—are developed using the weights generated by regression coefficient estimations. These two indexes and other linear function indexes are used as the predictors to predict corporate reputation, and their predictive validities are compared. Janis and Fadner (1943) argued that the weights of media coverage should be based on empirical evidence. This argument suggests a new perspective to develop new measures of media coverage, and find the solutions for the mixed findings that existed in the relationship between media coverage and corporate reputation. The current study is an attempt in such a direction.

The Exploration of the Relationship Between Media Coverage and Corporate Reputation

Media release much information about companies and actually play the role of information aggregators (Dyck & Zingales, 2002). Scholars in the management discipline have noticed the significant role of mass media in the business world and have been engaged in the exploration of this role for quite a long time. Fombrun and Shanley (1990) is the first prominent quantitative research exploring the relationship between media coverage and corporate reputation. They integrated media variables into the corporate reputation framework, and proposed three hypotheses, respectively, exploring the influence of media visibility, the tone of media coverage, and the interaction between them on corporate reputation at the overall level. However, they did not find any statistical support for these three hypotheses. Likewise, Wartick (1992) did not find the evidence to support the hypothesis that the amount and tone of media coverage have significant influence on corporate reputation at the overall level. However, he did find positive and significant correlations between media visibility and corporate reputation in the subgroups with “good” and “average” reputation levels. From a strategic resource perspective, Deephouse (2000) proposed a new concept: media reputation that is “the overall evaluation of a corporation presented in the media” (p. 1097).

Following these studies, communication scholars have explored the influence of media coverage on corporate reputation by applying the well-established media effects theories, especially the agenda-setting theory. Carroll and McCombs’ (2003) study is the first of such research studies in the communication discipline. They argued that although the agenda-setting theory primarily has been studied in political communication settings, “the central theoretical idea—the transfer of salience from the media agenda to the public agenda—fits equally well in the world of business communication” (p. 36). The agenda-setting theory theorizes the idea that mass media are stunningly successful in telling people what to think about and how to think about (McCombs, 2005). Scholars in this tradition identified two levels of agenda-setting effects: The first level of agenda setting argues the salience of an issue/object in media coverage has a direct effect on the salience of this issue/object in the public’s minds (Severin & Tankard, 2001). The second level of agenda setting holds that by emphasizing different attributes of an issue/object, and by expressing the positive, negative, or neutral tone of these attributes, media can not only influence the public’s perceptions of the salience of these attributes but also influence how they feel about this issue/object and its attributes (McCombs, 1997). These two levels of agenda setting have been verified in political news settings by many studies (e.g., Becker & McCombs, 1978; Kim & McCombs, 2007; McCombs, Llamas, Lopez-Excobar, & Rey, 1998; McCombs & Shaw, 1972). However, the agenda-setting theory has not received strong support from the empirical studies on the relationship between media coverage and corporate reputation.

Second-Level Agenda-Setting Effect of Business News

Second-level agenda setting is focused on the attribute characteristics and properties describing an issue/object. It has two dimensions: the cognitive dimension regarding the substantive (or cognitive) characteristics and the evaluative (or affective) dimension regarding the positive, negative, or neutral tone of these characteristics (McCombs & Reynolds, 2009). The cognitive dimension of second-level agenda setting presumes that various attributes/aspects of an issue/object may have different salience in news media, and this difference renders a similar pattern of salience of these attributes in public opinions. The affective dimension of second-level agenda setting states that the tone about an issue/object and its attributes described in news media will influence the public’s emotional perceptions of this issue/object and its attributes (McCombs & Ghanem 2001). Because corporate reputation is the assessment of consumers’ various emotional and cognitive perceptions of corporations (Ponzi, Fombrun, & Gardberg, 2011), second-level agenda setting with its two-dimensional framework is a powerful theoretical tool to investigate in depth the relationship between media coverage and corporate reputation.

In the examination of the relationship between media coverage and corporate reputation, scholars applied the characteristic attributes of corporate reputation to assess media coverage. For example, Kiousis et al. (2007) applied the characteristic attributes of The Harris Poll Reputation Quotient (RQ) to categorize corporate news into six attributes: vision and leadership, social responsibility, emotional appeal, products and services, workplace environment, and financial performance. Nevertheless, mixed findings were obtained in the previous studies. For example, Carroll (2004) found a significant correlation between media favorability and public esteem (the degree to which the public likes, trusts, and respects a corporation) at the overall level but not at the attribute level. Einwiller, Carroll, and Korn (2010) found a significant correlation between media favorability and corporate reputation in the social and environmental responsibility attribute. Meanwhile, Kiousis et al. (2007) did not find any significant correlation at either the overall level or the attribute level.

These mixed findings could be explained by different methodologies used in the previous studies. In particular, the mixed findings might be caused by the use of different news sample data. For example, some scholars employed The New York Times and The Wall Street Journal (Kiousis et al., 2007), whereas others used USA Today, and several major regional newspapers (Lee & Carroll, 2011). The industry sector might be another factor that contributes to the inconsistent findings (Fombrun & Shanley, 1990; Reputation Institute, 2012). The mixed findings might also be caused by the use of different measures of corporate reputation. For example, Kiousis et al. (2007) used the secondary data of Harris Poll RQ, whereas Meijer (2004) measured the first-hand corporate reputation ratings in her study. It should be highlighted that the previous studies used different measures of media coverage (Fombrun & Shanley, 1990; Kiousis, et al., 2007; Wry, Deephouse, & McNamara, 2006). Different measures could yield discrepant results in the measurement of media coverage and in the relationship between media coverage and corporate reputation. Thus, the discrepancy in the measurement of media coverage might also be one of the reasons for the mixed findings. The current study is focused on the measurement of media coverage and attempts to develop new measures that better predict corporate reputation.

Measures of Media Coverage of Corporations Using Linear Function

Scholars in both management and communication disciplines proposed multiple linear function models to measure the media coverage of corporations. Most of these measures are actually focused on media favorability. In particular, Kiousis et al. (2007) used the positive tone as the measure of media favorability. That is,

y = f,

where $f$ is the proportion of positive coding units and y is media favorability. This measure is called Kiousis index in the current study.

Fombrun and Shanley (1990) developed an index that measures the degree to which media reports are not negative. This media favorability index equals the sum of the proportions of positive ratings and neutral ratings. The formula is

y = f + n,

where $f$ is the proportion of positive coding units and $n$ is the proportion of neutral coding units. This measure is called Fombrun-Shanley index in the current study.

Wry et al. (2006) developed another linear function model to measure media favorability. Their media favorability index is the proportion of positive ratings minus the proportion of negative ratings. The formula is

y = f - u,

where $f$ is the proportion of positive coding units and $u$ is the proportion of negative coding units. This measure is called Wry index in the current study.

Meijer and Kleinnijenhuis (2006) developed a different measure that combines media favorability and media visibility. They assigned +1, +0.5, 0, −0.5, −1 to different tonalities, that is, positive, partially positive, neutral/mixed, partially negative, negative of the news articles, and summed the values as the measure of media coverage. That is

y = \sum_{1}^{n} t_{i},

where $t_{i}$ is the value of the tone of the news item $i$ , which is one of the numbers of +1, +0.5, 0, −0.5, −1. $n$ is the total number of the items. This measure is called Meijer-Kleinnijenhuis index in the current study.

Some studies argued that media visibility and media favorability are intertwined with each other (e.g., Kiousis, Bantimaroudis, & Ban, 1999; Zyglidopoulos & Gerorgiadis, 2006). Other studies demonstrated media visibility and its interaction with media favorability have significant influence on corporate reputation (e.g, Wartick, 1992; Fombrun & Shanley, 1990). Therefore, media visibility should not be excluded from the measurement of media coverage of corporations.

Media visibility plays different roles in various measures of media coverage. In particular, for the first three measures (Kiousis index, Fombrun-Shanley index, and Wry index), the number of positive/negative/neutral news units is the dividend, and the total number of news units is the divisor in their formulas. For Meijer & Kleinnijenhuis index, the number of positive/negative news units is the multiplier in the formula. As Fombrun and Shanley (1990) used the multiplication method to estimate the interaction effect of media favorability and visibility on corporate reputation, it is reasonable to argue that Meijer-Kleinnijenhuis index is a composite measure of media favorability and visibility. In this regard, Meijer-Kleinnijenhuis index is different from other three indexes that measure media favorability.

Moreover, because of the different roles of media visibility in the formulas, various indexes have different value ranges. Kiousis index and Fombrun-Shanley index have the value range [0, 1], whereas Wry index has the value range [−1, 1]. In a word, these three measures have the endpoints. Meanwhile, Meijer-Kleinnijenhuis index does not have endpoints. That is, the value of this index could be very large.

These measures are also different in the weights assigned to media tonalities. In particular, Wry index and Meijer-Kleinnijenhuis index assign the value of plus one (+1) to a positive unit of content, minus one (−1) to a negative unit, and zero (0) to a neutral unit. Fombrun-Shanley index assigns plus one (+1) to a positive unit as well as a neutral unit and zero (0) to a negative unit. Kiousis index assigns plus one (+1) to a positive unit and zero (0) to a negative unit as well as a neutral unit.

The assignment of plus one (+1) to a positive unit and minus one (−1) to a negative unit is based on the assumption that a piece of positive news has the same magnitude of impact as a piece of negative news does, but with opposing direction. However, this assumption is challenged by several studies that argued various tonalities of messages have different impacts in magnitude on an audience’s perceptions. For example, Mizerski (1982) demonstrated that negative information has more influence than positive information. Richey, Koenigs, Richey, and Fortin (1975) held that a single item of negative information is capable of neutralizing five similar pieces of positive information. These studies indicate that different values should be assigned to various tonalities of media content.

Research Questions and Hypotheses

Janis and Fadner (1943) argued that the numerical values assigned to media tonalities should be determined by empirical evidence. However, the assignments of tonality weights in the previous studies are discrepant, subjective, and lack empirical support. When various measures are used to examine the relationship between media coverage and corporate reputation, mixed results could be generated. Thus, it is necessary to calibrate the weights of media tonalities to address this problem. Therefore, the following research question is proposed:

Research Question 1 (RQ1): What weights should be assigned to tonalities in the measurement of media coverage of corporations at the overall level?

The relationship between media coverage and corporate reputation has also been investigated at the attribute level. The linear function measures can also be used to measure the media coverage of corporations in multiple attributes. Thus, the following research question is developed:

Research Question 2 (RQ2): What weights should be assigned to tonalities in the measurement of media coverage of corporations at the attribute level?

Linear regression analysis estimates the impacts of tonalities of media coverage on corporate reputation. The coefficients of linear regression model are the estimates of these impacts. Thus, these coefficients can be used as the weights of media tonalities to measure media coverage. The linear function measures in the previous studies do not differentiate neutral tonality and mixed tonality. Following this tradition, the linear regression model with three tonalities (positive, negative, and neutral/mixed) is developed as the follows:

C R_{i} = c + α f_{i} + β u_{i} + γ n_{i} + ε_{i},

where $C R_{i}$ is the corporate reputation score of firm i; $f_{i}$ is the proportion of positive coding units of firm i; $u_{i}$ is the proportion of negative coding units of firm i; $n_{i}$ is the proportion of neutral/mixed coding units of firm i; $α$ , $β$ , $γ$ are regression coefficients; c is the constant; and $ε_{i}$ is the residue.

When new weights are estimated by regression analysis, a new measure of media coverage of corporations can be developed. This new measure is called Three Factor Media Favorability (TFMF) index. The formula of this index is

T F M F i n d e x = α f + β u + γ n,

where $α$ , $β$ , $γ$ are regression coefficients of the regression model. $f$ is the proportion of positive coding units, $u$ is the proportion of negative coding units, and $n$ is the proportion of neutral/mixed coding units.

To estimate the coefficients ( $α$ , $β$ , $γ$ ), ordinary least square (OLS) method is used to minimize the sum of squared vertical distances between the observed responses and the responses predicted by the linear approximation. That is, these coefficients are generated through a process of data analysis. Thus, the weights of TFMF index are determined by regression algorithm.

The regression model (5) and the corresponding TFMF index (6) are based on the assumption that neutral and mixed tonalities are the same in predicting corporate reputation. Although this assumption has been applied in many previous studies (e.g., Fombrun & Shanley, 1990; Kiousis et al., 2007; Wry et al., 2006), Lee and Carroll (2011) differentiated the mixed tonality that has both positive and negative content from the neutral tonality that has neither positive nor negative content. This differentiation will affect the measurement of media coverage. To examine the effect of this differentiation, the following regression model is proposed:

C R_{i} = c + α f_{i} + β u_{i} + γ n_{i} + ω m_{i} + ε_{i},

where $C R_{i}$ is the corporate reputation score of firm i, $f_{i}$ is the proportion of positive coding units of firm i, $u_{i}$ is the proportion of negative coding units of firm i, $n_{i}$ is the proportion of neutral coding units of firm i, $m_{i}$ is the proportion of mixed coding units of firm i. $α$ , $β$ , $γ$ , $ω$ are regression coefficients, c is the constant, and $ε_{i}$ is the residue.

The coefficients of these four tonalities can be used as the weights to develop a new measure. This new measure is called Four Factor Media Favorability (FFMF) index. The formula of this index is

F F M F i n d e x = α f + β u + γ n + ω m,

where $α$ , $β$ , $γ$ , $σ$ , $ω$ are regression coefficients; $f$ is the proportion of positive coding units; $u$ is the proportion of negative coding units; $n$ is the proportion of neutral coding units; and $m$ is the proportion of mixed coding units.

Media visibility plays the same role in TFMF index and FFMF index as it does in Kiousis index, Fombrun-Shanley index, and Wry index. That is, the number of positive/negative/neutral/mixed news units is the dividend, and the total number of news units is the divisor. Thus, these two new indexes are also the measures of media favorability. And, they have the endpoints as well.

According to Janis and Fadner (1943), either TFMF index or FFMF index should have a higher predictive power than other measures because they are created by the empirically based weights instead of subjective weights. And, FFMF index should have a higher predictive power than TFMF index because the regression model of FFMF index has one more predictor than that of TFMF index does. Thus, the following hypotheses are proposed:

Hypothesis 1 (H1): TFMF index has a higher predictive power than other linear function indexes in predicting corporate reputation at the overall level.

Hypothesis 2 (H2): TFMF index has a higher predictive power than other linear function indexes in predicting corporate reputation at the attribute level.

Hypothesis 3 (H3): FFMF index has a higher predictive power than TFMF index and other linear function indexes in predicting corporate reputation at the overall level.

Hypothesis 4 (H4): FFMF index has a higher predictive power than TFMF index and other linear function indexes in predicting corporate reputation at the attribute level.

Method

Corporation Sample Selection

The analysis within a single industry enables the comparison of companies that compete directly with each other. The food industry was chosen as the sample industry in the current study because people are concerned with food information that may relate to their health. Furthermore, the information asymmetry problem is serious in food markets (Antle, 1999), which further increases the demand for food information. Some studies demonstrated that food information in news media has influence on consumers’ food knowledge and choices (Sheiham, Marmot, Rawson, & Ruck, 1987).

Reputation Institute (RI) is a company that consistently tracks the reputations of companies through extensive national surveys with large samples. The food companies on the ranking list of RI’s “Reputations of the 150 Largest Public U.S. Companies” comprise the sample of corporations in the current study. There are nine food companies in the sample, including General Mills, Kraft Foods, Kellogg, Sara Lee, H. J. Heinz, Dean Foods, ConAgra Foods, Tyson Foods, and Archer Daniels Midland. This is a small sample size. To increase the data size, the data of these nine companies within a 5-year period were analyzed.

Corporate Reputation Data

Following the previous studies conducted in the United States, the current study used the secondary data of corporate reputation ratings. Different from the previous studies that used The Harris Poll RQ (e.g., Carroll, 2004; Kiousis et al., 2007), the current study used the RepTrak® Pulse index developed by the RI.¹ The RepTrak® Pulse index is the assessment of the various emotional and cognitive perceptions of corporate reputation (Ponzi et al., 2011). Meanwhile, emotional appeal is only one of the six attributes in The Harris Poll RQ (The Harris Poll, 2013). In this regard, the RepTrak® Pulse index is superior to the Harris Poll RQ for examining the second-level agenda-setting effect of corporate news. The corporate reputation scores of the sample companies at the overall level and attribute level from 2008 to 2012 were obtained from the RI.

News Sample Selection

The news population of the current study consists of both the elite newspapers and local newspapers. Because The New York Times has been a benchmark for media agenda (Gans, 1979), and has the ability of inter-media agenda setting (Gilbert, Eyal, McCombs, & Nicholas, 1980; Reese & Danielian, 1989; Shaw & Sparrow, 1999), it was chosen as one of the elite newspapers. The Wall Street Journal was also chosen as another elite newspaper because it has significant influence on business news (Kiousis et al., 2007). Local newspapers were included in the news population because the Pew’s national survey showed that local newspapers are the major information sources for most Americans (Pew Research Center, 2012).

News articles published in the time period from 2007 to 2011 constitute the sampling frame of news population. Because The Wall Street Journal news data cannot be obtained from Lexis-Nexis, Factiva was the database used to obtain news articles in the two elite newspapers. Because Lexis-Nexis contains larger number of newspapers than Factiva, it was used as the source for local newspapers. The search procedure of the current study generated a local newspaper population that contains many major metro newspapers,² such as The Washington Post and The Chicago Tribune.

Lexis-Nexis and Factiva both have a search tool that helps locate the central theme of news articles: search the keywords in the headline and lead paragraph. This search tool was used as a filter to search the names of corporations. However, the two databases have different search tools for screening the search results for specific purposes. In particular, Lexis-Nexis has the “ticker” tool, which is the code used to uniquely identify a publicly traded company on a stock market. This tool was used as a filter to identify a corporation’s presence in the article. This search tool was combined with the keywords (corporations’ names) tool to search news articles about the corporations. Factiva does not have the “ticker” filtering tool. Instead, it has the industry filter (food + beverages/tobacco), which was used as the alternative tool to screen news articles. This filter was combined with the keywords (corporations’ names) tool to search news articles about the corporations. In the second stage of searching, the brand names belonging to these corporations but different from the corporations’ names were also used as the keywords for search. Another two filter tools, ticker and industry filter, were also combined with the keywords filter to search news articles about the brands. In order to avoid the duplication of the news articles in news population, the articles containing corporations’ names in the keywords were excluded from this stage of search.

All the news articles obtained by these search methods were downloaded. Then, several types of irrelevant articles were filtered out manually. Articles falling into the following categories were excluded from the final news population: story that does not refer to the corporation/brand, corrections/clarification, obituary, news forecast with the headlines “Looking Ahead” and “Coming Up,” and duplicate articles.

Through these search and filtering processes, 3,396 news articles were obtained among which 260 from The New York Times, 412 from The Wall Street Journal, and 2,724 from local newspapers. News sample articles were drawn from this news population. The news sample size of each company in each year was calculated with 95% confidence level and 5% margin of error.³ This calculation was performed for The New York Times, The Wall Street Journal, and local newspapers separately. For example, in the news population of 2007, General Mills has 163 news articles (12 from The New York Times, 16 articles from The Wall Street Journal, and 135 from local newspapers). The sample size calculation results are: 12 for The New York Times, 15 for The Wall Street Journal, and 100 for local newspapers. Thus, in the news sample, General Mills has 127 news articles. The total sample size of the nine companies in the time period from 2007 to 2011 is the sum of the sample size of each company in each year, which is 2,817 (255 for The New York Times, 397 for The Wall Street Journal, and 2,165 for local newspapers). The numbers of news articles in the news sample for each company from the two elite newspapers and local newspapers are shown in Table 1. These numbers of news articles were randomly selected from the news population.

Table 1.

Number of Articles of Nine Food Companies in the Elite and Local Newspapers.

	NYT	NYT	NYT	NYT	NYT	WSJ	WSJ	WSJ	WSJ	WSJ	Local	Local	Local	Local	Local
	2007	2008	2009	2010	2011	2007	2008	2009	2010	2011	2007	2008	2009	2010	2011
ADM
Frequency	3	3	3	2	3	9	11	5	3	1	29	33	50	25	36
%	6.12	8.82	5.17	3.51	5.26	10.00	19.64	5.43	3.19	1.54	9.35	12.04	8.24	4.94	7.69
CAG
Frequency			7	5	6			3	11	9			55	36	42
%			12.07	8.77	10.53			3.26	11.70	13.85			9.06	7.11	8.97
DF
Frequency			2	1	0			2	5	4			27	30	24
%			3.45	1.75	0.00			2.17	5.32	6.15			4.45	5.93	5.13
GIS
Frequency	12	4	12	7	9	15	9	10	11	11	100	96	117	91	105
%	24.49	11.76	20.69	12.28	15.79	16.67	16.07	10.87	11.70	16.92	32.26	35.04	19.28	17.98	22.44
HNZ
Frequency			5	3	3			4	4	4			47	58	51
%			8.62	5.26	5.26			4.35	4.26	6.15			7.74	11.46	10.90
K
Frequency			6	10	6			10	8	2			95	56	61
%			10.34	17.54	10.53			10.87	8.51	3.08			15.65	11.07	13.03
KFT
Frequency	16	11	17	19	14	36	10	36	34	16	96	73	135	123	74
%	32.65	32.35	29.31	33.33	24.56	40.00	17.86	39.13	36.17	24.62	30.97	26.64	22.24	24.31	15.81
SLE
Frequency	8	3	3	5	9	13	11	11	11	12	23	14	34	42	44
%	16.33	8.82	5.17	8.77	15.79	14.44	19.64	11.96	11.70	18.46	7.42	5.11	5.60	8.30	9.40
TSN
Frequency	10	13	3	5	7	17	15	11	7	6	62	58	47	45	31
%	20.41	38.24	5.17	8.77	12.28	18.89	26.79	11.96	7.45	9.23	20.00	21.17	7.74	8.89	6.62
Total	49	34	58	57	57	90	56	92	94	65	310	274	607	506	468

Note. The names of the companies are abbreviated by their tickers. In particular, ADM = Archer Daniels Midland; CAG = ConAgra Foods; DF = Dean Foods; GIS = General Mills; HNZ = H. J. Heinz; K = Kellogg; KFT = Kraft Foods; SLE = Sara Lee; TSN = Tyson Foods. The New York Time is abbreviated by NYT, The Wall Street Journal is abbreviated by WSJ, and local newspapers are abbreviated by Local. The news reports of CAG, DF, HNZ, and K are not included in the sample in 2007 and 2008 because the reputation measurement data of these firms are not available. The percentage (%) number in the table is calculated by the number of articles of a company divided by the total number of articles of the nine companies in NYT, WSJ, and Local newspapers in each year.

Content Analysis

A pilot coding was conducted to train coders and build consensus at the earlier stage. The codebook and coding sheet were developed at this stage. Then, the author and another coder worked together to code 300 news articles. The first five paragraphs were read and the coders used the information in these paragraphs to answer the coding questions. When the coders felt the first five paragraphs did not provide sufficient information, they read a few more paragraphs. The codebook had been revised continually during the training process so that the coders’ coding approaches were calibrated. In the next coding stage, 300 news articles, which constitute 10.2% of the articles in the news sample, were randomly selected and coded independently, without any discussion or collaboration. Blind coding technique, which requires that another coder does not know the purpose of the research, was applied in the coding process (Neuendorf, 2002).

The current study uses the characteristic attributes of the RI’s RepTrak® Pulse index to categorize news content into seven attributes. During the coding process, the coders identified one or several attributes in the news articles. The brief definitions of these seven attributes are listed as follows: (1) products or services, which relates to the perceptions of the quality, value, and reliability of a company’s products and services; (2) leadership, which relates to how much a company demonstrates a clear vision and strong leadership; (3) financial performance, which relates to the perceptions of a company’s profitability, prospects, and risk; (4) innovation, which refers to how a company makes or sells innovative products or innovates in the way it does business; (5) citizenship, which relates to the perceptions of a company as a good citizen in its dealings with communities, employees, and environment; (6) governance, which refers to whether a company behaves ethically, and is open and transparent in its business dealings; and (7) workplace, which relates to the perceptions of how well a company is managed, how it is to work for, and the quality of its employees (Fombrun, 2006).

Each article was coded into different tonalities: positive, negative, neutral, and mixed. According to Lee and Carroll (2011), positive news refers to the content that indicates a company has positive emotional appeal, is an object of admiration and respect, or particularly trustworthy. Negative news refers to the content indicating a company generates negative emotional appeal, or is portrayed as unworthy of admiration, respect, or trust. Neutral news refers to the content that does not have any positive and negative connotations. Mixed news refers to the content that relates to a company that has both positive and negative connotations in the story. However, Pollock and Rindova (2003) defined neutral news as the news articles with relatively equal instances of positive and negative references. To code different tonalities, the current study uses the definitions of positive, negative, and neutral news proposed by Lee and Carroll (2011), and uses the definition of neutral news proposed by Pollock and Rindova (2003) for mixed news. The tonality coding was also conducted at the attribute level. That is, once one of the seven attributes was identified, its tonality was coded as one of the four tonalities.

The inter-coder reliability was computed for the coding work. The minimum acceptable level of the inter-coder reliability coefficients is selected as .80, suggested by Lombard, Snyder-Duch, and Bracken (2002). Following the previous studies (Kiousis et al., 2007; Ragas, 2010), two inter-coder reliability indexes, Holsti’s (1969) formula and Scott’s Pi, were examined. The coding and inter-coder reliability test processes were repeated until the values of both indexes reach the level of .80 or higher for all variables in the coding sheet.

In the news sample, there are some articles in which only the name of a focal company is mentioned, but the text provides no relevant information about that company. These articles were coded as irrelevant articles. There were 26 irrelevant news articles from The New York Times, 23 irrelevant news articles from The Wall Street Journal, and 274 irrelevant news articles from the local newspapers. Once the irrelevant articles were identified, their tonalities and attributes were not coded anymore.

The Fixed Effects Regression Model With Panel Data

The data used in the current study consist of five consecutive year data of nine food companies. They are panel data because the data set is “constructed from repeated cross sections over time” (Wooldridge, 2003, p. 842). As the corporate reputation ratings of four food companies (ConAgra Foods, Dean Foods, H. J. Heinz, and Kellogg) were not measured by RI in 2007 and 2008, the data set of the current study does not contain these data, resulting in an unbalanced panel with 37 observations. Panel data model can be used to control for variables that cannot be observed or measured, and can be used for variables at different levels of entities, such as companies (Baltagi, 2008). It is an appropriate statistical method for the current study.

Wooldridge (2003) stated that the fixed effects model should be a better choice than the random effects model when the observations are not randomly drawn from a large population. Because the sample companies are the nine food companies in the list of RI’s “Reputations of the 150 Largest Public U.S. Companies,” they are not randomly drawn from a large firm population. Moreover, the company-specific intercept α_i has a specific implication in the current study: a company’s reputation level when its news coverage is zero. It reflects the unobserved effect of all other variables that could influence corporate reputation. These variables include profitability, risk, advertising, size, charity, foundation, diversification, and so forth (Fombrun & Shanley, 1990). These variables could influence media coverage (Carroll & McCombs, 2003), and thus, the unobserved effect α_i is correlated with media coverage. In such conditions, fixed effects model is a more appropriate analytical tool than the random effects model (Wooldridge, 2003). The equation of the fixed effects model is

Y_{i t} = α_{i} + β^{’} x_{i t} + u_{i t}, t = 1, 2, . . ., T,

where Y_it is the dependent variable (DV), i = entity, and t = time; α_i is the intercept for each entity; β^’ is the coefficient for independent variable(s) (IV); x_it is IV(s); u_it is the residue.

Results

RQ1 asks what weights should be assigned to tonalities to measure media coverage of corporations at the overall level. The coefficients of regression model can be used as the weights because they represent the impacts of these tonalities on corporate reputation. Table 2 shows the estimations of the coefficients of tonalities for TFMF index and FFMF index. For TFMF index, positive tonality has a positive coefficient (.023) and negative tonality has a negative coefficient (−.127) for the elite newspapers; positive tonality has a positive coefficient (.045) and negative tonality has a negative coefficient (−.286) for the local newspapers. And the absolute values of the coefficient of negative tonality are much larger than those of positive tonality for both the elite newspapers and local newspapers. Neutral/mixed tonality has a negative coefficient (−.018) for the elite newspapers, whereas it has a positive coefficient (.111) for the local newspapers. For FFMF index, positive tonality and negative tonality have the same coefficients as TFMF index. Neutral tonality has a positive coefficient (.178) and mixed tonality has a negative coefficient (−.256) for the elite newspapers. And neutral tonality has a positive coefficient (.166), mixed tonality has a negative coefficient (−.252) for the local newspapers.

Table 2.

Regression Estimates of the Coefficients of Tonalities in Predicting Corporate Reputation Score at the Overall Level.

	NWPP	NWNP	NWUM	NWUP	NWMP	LOCPP	LOCNP	LOCUM	LOCUP	LOCMP
TFMF index
Coefficients	.023	−.127	−.018			.045	−.286*	.111
Significance	.767	.133	.863			.676	.011	.314
FFMF index
Coefficients	.023	−.127		.178	−.256*	.045	−.286*		.166	−.252
Significance	.767	.133		.065	.019	.676	.011		.089	.156

Note. NWPP is the percentage of positive articles in the elite newspapers; NWNP is the percentage of negative articles in the elite newspapers; NWUM is the percentage of neutral and mixed articles in the elite newspapers; NWUP is the percentage of neutral articles in the elite newspapers; NWMP is the percentage of mixed articles in the elite newspapers; LOCPP is the percentage of positive articles in the local newspapers; LOCNP is the percentage of negative articles in the local newspapers; LOCUM is the percentage of neutral and mixed articles in the local newspapers; LOCUP is the percentage of neutral articles in the local newspapers; LOCMP is the percentage of mixed articles in the local newspapers. Newspaper sample data from 2007 to 2011 were coded and used to calculate the proportions of tonalities. RepTrak^@ Pulse data from 2008 to 2012 were used as the dependent variable. Fixed effect model was used for estimation. Number of observation n = 37.

p < .05.

RQ2 asks what weights should be assigned to tonalities to measure media coverage of corporations at the attribute level. Table 3 shows the estimations of the coefficients of tonalities for TFMF index. Different from the results of the estimations at the overall level, the coefficients of tonalities are mixed in directions. For example, positive tonality has a positive coefficient (.199), negative tonality also has a positive coefficient (.131), and neutral/mixed tonality has a negative coefficient (−.143) in the products and services attribute for the elite newspapers; positive tonality has a positive coefficient (.207), negative tonality has a negative coefficient (−.176), and neutral/mixed tonality has a positive coefficient (.004) in the products and services attribute for the local newspapers. For the elite newspapers, only one attribute (financial performance) is found whose coefficients of positive tonality and negative tonality have consistent directions with those at the overall level. For the local newspapers, five of seven attributes (products and services, financial performance, citizenship, governance, and workplace) are found whose coefficients of positive tonality and negative tonality have consistent directions with those at the overall level.

Table 3.

Regression Estimates of the Coefficients of Three Tonalities in Predicting Corporate Reputation Score at the Attribute Level.

	NWPP	NWNP	NWUM	LOCPP	LOCNP	LOCUM
Products and services
Coefficients	.199	.131	−.143	.207*	−.176	.004
Significance	.144	.179	.059	.048	.063	.056
Leadership
Coefficients	−.025	−.011	.031	−.006	.158	−.009
Significance	.822	.919	.259	.868	.171	.799
Financial performance
Coefficients	.008	−.054	.035	.035	−.014	−.021
Significance	.830	.275	.238	.352	.738	.557
Innovation
Coefficients	.033*	—	—	−.015	.035	.038
Significance	.041	—	—	.548	.817	.570
Citizenship
Coefficients	.031	.090	−.061	.036	−.059	−.071
Significance	.221	.161	.191	.921	.271	.429
Governance
Coefficients	.024	.011	.013	.138**	−.053	.031
Significance	.551	.358	.665	.002	.119	.373
Workplace
Coefficients	−.013	.016	−.014	.068*	−.064*	−.016
Significance	.737	.587	.672	.010	.023	.634

Note. For the elite newspapers, no negative and neutral/mixed report regarding innovation was found in the news sample. Thus, the coefficients of these tonalities were not obtained. Newspaper sample data from 2007 to 2011 were coded and used to calculate the proportions of tonalities. RepTrak^@ Pulse data from 2008 to 2012 were used as the dependent variables. Fixed effect model was used for estimation. Number of observation n = 37.

p < .05. **p < .01.

Table 4 shows the estimations of the coefficients of tonalities for FFMF index. The coefficients of positive tonality and negative tonality are the same as those of TFMF index. The coefficients of neutral tonality and mixed tonality are mixed in directions for both the elite newspapers and local newspapers.

Table 4.

Regression Estimates of the Coefficients of Four Tonalities in Predicting Corporate Reputation Score at the Attribute Level.

	NWPP	NWNP	NWUP	NWMP	LOCPP	LOCNP	LOCUP	LOCMP
Products and services
Coefficients	.199	.131	−.143	—	.207*	−.176	−.019	.265
Significance	.144	.179	.059	—	.048	.063	.780	.254
Leadership
Coefficients	−.025	−.011	.042	−.067	−.006	.158	.000	−.107
Significance	.822	.919	.199	.372	.868	.171	.200	.380
Financial performance
Coefficients	.008	−.054	.047	−.092	.035	−.014	.003	−.094
Significance	.830	.275	.244	.274	.352	.738	.936	.178
Innovation
Coefficients	.033*	—	—	—	−.015	.035	.085	.032
Significance	.041	—	—	—	.548	.817	.564	.716
Citizenship
Coefficients	.031	.090	.010	−.112	.036	−.059	−.246	−.059
Significance	.221	.161	.882	.063	.921	.271	.353	.569
Governance
Coefficients	.024	.011	−.097	.041	.138**	−.053	.022	.034
Significance	.551.	.358	.120	.200	.002	.119	.615	.490
Workplace
Coefficients	−.013	.016	.022	−.087	.068*	−.064*	.031	−.049
Significance	.737	.587	.555	.107	.010	.023	.509	.238

Note. For the elite newspapers, no mixed report regarding products and services was found; no negative, neutral, and mixed report regarding innovation was found in the news sample. Thus, the coefficients of these tonalities were not obtained. Newspaper sample data from 2007 to 2011 were coded and used to calculate the proportions of tonalities. RepTrak^@ Pulse data from 2008 to 2012 were used as the dependent variable. Fixed effect model was used for estimation. Number of observation n = 37.

p < .05. **p < .01.

Two new measures (FFMF index and TFMF index) of media favorability are developed using the coefficients of different tonalities as the weights. These two new measures, as well as other linear function measures, are used as the predictors to predict corporate reputation. Table 5 shows the results of these regressions.

Table 5.

The Comparison of the Prediction Power of Media Coverage Indexes at the Overall and Attribute Level.

	Index	Overall R²	Products and services R²	Leadership R²	Financial performance R²	Innovation R²	Citizenship R²	Governance R²	Workplace R²
Elite newspapers	Kiousis index	.001	.06	<.001	.012	.114	.042	.007	.003
	Fombrun-Shanley index	.083	.051	.028	.079	.114	.048	.003	.001
	Wry index	.02	.002	<.001	.019	.114	.01	<.001	.011
	Meijer-Kleinnijenhuis index	.141	.007	.072	.057	.176	.057	.001	.016
	TFMF index	.051	.101	.014	.053	.114^a	.138	.03	.015
	FFMF index	.22	.101	.049	.066	.114	.177	.128	.091
Local newspapers	Kiousis index		.108	.001	.025	.01	.024	.246	.176
	Fombrun-Shanley index	.213	.036	.004	.032	.007	.018	.179	.175
	Wry index	.061	.176	.009	.017	.011	.035	.205	.228
	Meijer-Kleinnijenhuis index	.224	.099	.145	.077	.125	.196	.108	.119
	TFMF index	.165	.177^a	.052	.052	.017	.047	.272^a	.235^a
	FFMF index	.181	.238 ^a	.07	.07	.019	.057	.274 ^a	.263 ^a

Note. The number in bold font represents the highest value in comparison with others. Fixed effect model was used for estimation. Number of observation n = 37.

The R² value of the measure (TFMF or FFMF) created by the weights using significant regression coefficients.

H1 examines whether TFMF index has higher a predictive power than other linear function indexes in predicting corporate reputation at the overall level. It does not receive support because the predictive power (R² = .051) of TFMF index for the elite newspapers is lower than that (R² = .083) of Fombrun-Shanley index and that (R² = .141) of Meijer-Kleinnijenhuis index, and the predictive power (R² = .165) of TFMF index for the local newspapers is lower than that (R² = .213) of Fombrun-Shanley index and that (R² = .224) of Meijer-Kleinnijenhuis index.

H2 examines whether TFMF index has a higher predictive power than other linear function indexes in predicting corporate reputation at the attribute level. It received partial support because the predictive power of TFMF index for the elite newspapers is higher than those of other indexes in three attributes: (1) products and services (R² = .101), (2) citizenship (R² = .138), and (3) governance (R² = .030). Moreover, the predictive power of TFMF index for the local newspapers is higher than those of other indexes in three attributes: (1) products and services (R² = .177), (2) governance (R² = .272), and (3) workplace (R² = .235).

H3 examines whether FFMF index has a higher predictive power than TFMF index and other linear function indexes in predicting corporate reputation at the overall level. It receives support for the elite newspapers because the predictive power (R² = .220) of FFMF index is higher than that (R² = .051) of TFMF index and those of other indexes. Meanwhile, it does not receive support for the local newspapers because the predictive power (R² = .181) of FFMF index is lower than that (R² = .213) of Fombrun-Shanley index and that (R² = .224) of Meijer-Kleinnijenhuis index.

H4 examines whether FFMF index has a higher predictive power than TFMF index and other linear function indexes in predicting corporate reputation at the attribute level. It receives partial support for the elite newspapers because the predictive power of FFMF index is higher than that of TFMF index and other indexes in four attributes: (1) products and services (R² = .101), (2) citizenship (R² = .177), (3) governance (R² = .128), and (4) workplace (R² = .091). It also receives partial support for the local newspapers because the predictive power of FFMF index is higher than that of TFMF index and other indexes in three attributes: (1) products and services (R² = .238), (2) governance (R² = .274), and (3) workplace (R² = .263).

Discussion

Second-level agenda setting argues that news media not only influence the public’s perceptions of the salience of attributes of an issue/object but also influence how they feel about this issue/object and its attributes (McCombs, 1997). This theory has been used as the theoretical foundation to investigate the relationship between media coverage and corporate reputation. However, scholars developed various measures of media coverage to examine this relationship (e.g., Carroll, 2004; Fombrun & Shanley, 1990; Kiousis et al., 2007; Wry et al., 2006). As mixed findings were obtained by these studies, it is necessary to examine whether the difference in the measurement of media coverage renders this problem.

The current study explores what weights of media tonalities should be used in the measurement of media coverage. Based on the argument that weights assigned to media tonalities should be determined by empirical evidence (Janis & Fadner, 1943), the current study estimates the weight values by using linear regression analysis. The regression coefficient estimations clearly show negative tonality has much stronger influence on corporate reputation than positive tonality at the overall level. This finding, in line with the research results obtained by Fan, Geddes, and Flory (2013), echoes the findings of some previous studies that showed positive and negative information has different magnitudes of influence on people’s perceptions (e.g., Mizerski, 1982; Richey et al., 1975). Wry index and Meijer-Kleinnijenhuis index assign positive tonality and negative tonality with the same absolute value but opposing directions. The findings of the current study suggest that the absolute values assigned to positive tonality and negative tonality should also be different.

The measures analyzed in the current study evaluate different aspects of media coverage as media visibility plays different roles in their formulas. In particular, Kiousis index, Fombrun-Shanley index, Wry index, TFMF index, and FFMF index measure media favorability, whereas Meijer-Kleinnijenhuis index is a composite measure of media favorability and visibility. A media favorability measure is appropriate to examine the relationship between media favorability and the public esteem of corporations (Carroll, 2004). Meanwhile, a composite measure of media favorability and visibility is appropriate to examine the interaction effect of these two factors on corporate reputation (Fombrun & Shanley, 1990).

The comparison of five measures of media favorability clearly reveals the improvement of measurement by using the empirically based weights. TFMF index and FFMF index have advantages over other three media favorability measures especially at the attribute level. In particular, TFMF index has higher predictive power than those of other three indexes in four attributes (products and services, citizenship, governance, and workplace) for the elite newspapers. And, it has higher predictive power than those of other three indexes in all seven attributes for the local newspapers. FFMF index has higher predictive power than other three indexes in five attributes (products and services, leadership, citizenship, governance, and workplace) for the elite newspapers. And, it has higher predictive power than other three indexes in all seven attributes for the local newspapers.

It is obvious that FFMF index is superior to TFMF index in predicting corporate reputation. The difference between these two measures is that FFMF index differentiates neural tonality and mixed tonality, whereas TFMF index takes them as equal. The superiority of FFMF index supports Lee and Carroll’s (2011) argument that neutral and mixed tonalities should be differentiated in the measurement of media favorability.

Nevertheless, the comparison also shows neither TFMF index nor FFMF index has a complete superiority in predicting corporate reputation. For example, Fombrun-Shanley index has a higher predictive power than the two new measures in predicting financial performance attribute reputation for the elite newspapers, and in predicting overall corporate reputation for the local newspapers. These inconclusive findings suggest that the improvement in the measurement of media favorability is not sufficient to build a measure that has a complete superiority in predicting corporate reputation at the overall level and attribute level.

The comparison between media favorability measures and a composite measure of media favorability and visibility reveals more insights. Compared with five media favorability indexes, Meijer-Kleinnijenhuis index has a higher predictive power at the overall level for the local newspapers, has a higher predictive power in two attributes (leadership and innovation) for the elite newspapers, and has a higher predictive power in four attributes (leadership, financial performance, innovation, and citizenship) for the local newspapers. These findings suggest the composite measure of media favorability and visibility has relative advantages in predicting corporate reputation over media favorability measures. The composite measure actually reflects the interaction between media favorability and visibility. Fombrun and Shanley (1990) demonstrated that this interaction has stronger correlation with corporate reputation than media favorability does.

It should be noticed that no measure has the predictive power (R²) higher than .30. Media coverage is only one of the factors that could influence corporate reputation. Many other factors, such as profitability, risk, advertising, size, charity, foundation, and diversification also influence corporate reputation (Fombrun & Shanley, 1990). Thus, it is unrealistic to create a measure of media coverage that perfectly predicts corporate reputation.

The previous studies used different news sources to measure media coverage of corporations (e.g., Kiousis et al., 2007; Lee & Carroll, 2011). Different news sources could generate discrepant results in the measurement of media coverage. The current study finds that the coefficients of five attributes have consistent directions with those at the overall level for the local newspapers, and the coefficients of one attribute have consistent directions with those at the overall level for the elite newspapers. Moreover, more coefficients are significant for the local newspapers than for the elite newspapers at either the overall level or attribute level. These findings suggest that local newspapers are better news sources than elite newspapers to estimate the coefficients of media tonalities.

It is also found that different news samples render discrepant outcomes in predicting corporate reputation. The local newspaper sample has more news reports about the sample companies as it consists of many more newspapers than the elite newspaper sample. Corporate reputation is relatively stable (Walker, 2010). That is, the reputation of a company should not change drastically in a short time period. Meanwhile, if only a few news articles are used to measure media coverage of a company, the results would be unstable. For instance, a company has one positive report this year, and one negative report next year. Its media coverage evaluation will change drastically when one media favorability measure is used. This drastic change would happen more likely when a small number of news reports are used than when a large number of news reports are used. Thus, a threshold of media visibility is needed to create a valid measure of media favorability. Of course, hundreds of local newspapers have more news reports on a company than a few elite newspapers. Because of the need of a threshold of media visibility, local newspapers are better news sources than elite newspapers to measure media favorability.

As the first study testing the weights of media tonalities of corporations, the current study has several limitations. The first limitation is the generalizability of the findings. The current study is focused on the food industry. The Reputation Institute (2012) argued that consumers’ perceptions about an industry have a halo effect on an individual company within that industry. This provides the rationale for the research study within one industry. However, as Schudson (1984) argued, media as well as other information sources may have different degrees of influence on people’s attitudes toward various products/services. People care about food information that relates to their health. They may pay more attention to food messages than to other messages. And, media coverage of food companies may have greater influence on people’s perceptions than that of other types of companies. Thus, some cautions should be taken when generalizing the conclusions of the current study to other industries.

Another limitation is the small number of companies in the sample. Only nine companies were included in the corporation sample and only 37 observations were obtained. The small number of observations limits the application of multivariate analysis. More sample companies or more time periods should be included in future research in order to apply multivariate analysis. Then, the number of sample news articles would be greatly amplified. Due to the limitation of time and other resources, human coding method is not as good as the computer-aided text analysis (CATA) to do content analysis for such big data.

It should be noticed that the coefficients of tonalities of media coverage exhibit a chaotic pattern and many of them are not significant, as shown in Table 3 and Table 4. This chaotic pattern might be caused by multiple factors. The multicollinearity among the measures of tonalities might be one of the reasons, although a full multicollinearity does not occur because of the existence of irrelevant news articles. The different impacts of various media content on people’s perceptions might be another reason. Zhang (2014) argued that news messages of different attributes (e.g., products and services, leadership, financial performance) have different influences on people’s perceptions toward the corresponding attribute reputations. This argument helps explain the differences of the coefficients across attributes. The small news sample size at the attribute level would be another reason for these non-significant findings, as more significant coefficients are obtained for the local newspapers than for the elite newspapers. The CATA would help solve this problem if the software is developed to code the attribute information in the text.

The data used in the current study also have limitations. The data of the dependent variable—corporate reputation scores—were obtained from the RI, and thus carry the limitations of secondary data. The author has no control over the sampling methods, survey questionnaires and techniques, and the methodologies used for calculating corporate reputation scores. The news data were obtained from two large news databases: Lexis-Nexis and Factiva. Although they are very large databases, there are still some newspapers/articles that are not included (Weaver & Bimber, 2008). And, some relevant articles may not be obtained through the search tools provided by these data sets.

The current study is focused on the measures using linear function formulas. Scholars also developed nonlinear function formulas to measure media content. For example, Lowe, Benoit, Mikhaylov, and Laver (2011) developed a logit scale to measure the content of party manifestoes. Pollock (1995) developed a nonlinear composite measure of media favorability and visibility to evaluate media coverage of social issues, such as community/city/nation characteristics, social change, social inequality, and human rights. Future research should use these nonlinear formulas to measure media coverage of corporations, and explore the differences between linear measures and nonlinear measures. More sophisticated statistical analyses should be conducted on these measures, such as variance/volatility (the second moment), skewness (the third moment), and Kurtosis (the fourth moment).

In closing, the current study develops two new measures with the weights of tonalities estimated by regression analysis and compares them with other linear function measures of media coverage. The findings suggest a certain degree of relative advantage of these two new measures in predicting corporate reputation, and a certain degree of relative advantage of a composite measure of media favorability and visibility in predicting corporate reputation. These findings provide new insights into the measurement of media coverage of corporations and help researchers further develop new measures with higher advantages. The advance in this direction would further the exploration of media effects in the business news settings. Media effects research in other news settings would also benefit from this advance.

Footnotes

Acknowledgements

The author would like to thank Dr. Michael Elwood Roloff and the anonymous reviewers for their constructive comments and suggestions.

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biography

Xiaoqun Zhang gained his first PhD in management from Tsinghua University, and his second PhD in media studies from Bowling Green State University. His research interests include media management and economics, media technology and policy, and the measurement of media coverage. His work has been recently published in Telecommunications Policy, Journalism & Mass Communication Quarterly, Time & Society, and Corporate Reputation Review.

References

Antle

(1999). The new economics of agriculture. American Journal of Agricultural Economics, 81, 993-1010. doi:10.2307/1244078

Baltagi

B. H.

(2008). Econometric analysis of panel data (4th ed.). Hoboken, NJ: John Wiley.

Becker

L. B.

McCombs

M. E.

(1978). The role of the press in determining voter reactions to presidential primaries. Human Communication Research, 4, 301-307.

Carmines

E. G.

Zeller

R. A.

(1979). Reliability and validity assessment (Vol. 17). Newbury Park CA: Sage. doi:10.4135/9781412985642

Carroll

C. E.

(2004). How the mass media influence perceptions of corporate reputation: Exploring agenda-setting effects within business news coverage (Unpublished doctoral dissertation). University of Texas at Austin, TX.

Carroll

C. E.

McCombs

M. E.

(2003). Agenda-setting effects of business news on the public’s images and opinions about major corporations. Corporate Reputation Review, 6, 36-46. doi:10.1057/palgrave.crr.1540188

Davis

(2000). Public relations, news production and changing patterns of source access in the British national press. Media, Culture & Society, 22, 39-59. doi:10.1177/016344300022001003

Deephouse

D. L.

(2000). Media reputation as a strategic resource: An integration of mass communication and resource-based theories. Journal of Management, 26, 1091-1112. doi:10.1177/014920630002600602

Dyck

Zingales

. (2002). The corporate governance role of media. In World Bank Institute (Ed.), The right to tell: The role of mass media in economic development (pp. 107-140). Washington, DC: World Bank Publications.

10.

Einwiller

Carroll

C. E.

Korn

(2010). Under what conditions do the news media influence corporate reputation? The roles of media systems dependency and need for orientation. Corporate Reputation Review, 12, 299-315. doi:10.1057/crr.2009.28

11.

Fan

Geddes

Flory

(2013). The Toyota recall crisis: Media impact on Toyota’s corporate brand reputation. Corporate Reputation Review, 16, 99-117. doi:10.1057/crr.2013.6

12.

Fombrun

C. J.

(2006, May 25-28). The RepTrak system. Presented at 10th Anniversary Conference on Reputation, Image, Identity and Competitiveness, New York, NY.

13.

Fombrun

C. J.

Shanley

(1990). What’s in a name? Reputation building and corporate strategy. Academy of Management Journal, 33, 233-258. doi:10.2307/256324

14.

Gans

H. J.

(1979). Deciding what’s news: A study of CBS evening news, NBC nightly news Newsweek and Time. New York, NY: Pantheon.

15.

Gilbert

Eyal

McCombs

M. E.

Nicholas

(1980). The state of the union address and press agenda. Journalism Quarterly, 57, 584-588. doi:10.1177/107769908005700405

16.

Henriques

D. B.

(2000). Business reporting: Behind the curve. Columbia Journalism Review, 39(4), 18-21.

17.

Holsti

(1969). Content analysis for the social sciences and humanities. Don Mills, Ontario, Canada: Addison-Wesley.

18.

Janis

I. L.

Fadner

(1943). The coefficient of imbalance. Psychometrica, 8, 105-119. doi:10.1007/BF02288695

19.

Jones

(1987). The media and industrial relations: The changing relationship. Warwick Papers in Industrial Relations, 18, 1-14.

20.

Kim

McCombs

M. E.

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

21.

Kiousis

Bantimaroudis

Ban

. (1999). Candidate image attributes: Experiments on the substantive dimension of second level agenda setting. Communication Research, 26, 414-428. doi:10.1177/009365099026004003

22.

Kiousis

Popescu

Mitrook

(2007).Understanding influence on corporate reputation: An examination of public relations efforts, media coverage, public opinion, and financial performance from an agenda-building and agenda-setting perspective. Journal of Public Relations Research, 19, 147-165. doi:10.1080/10627260701290661

23.

Krippendorff

(1980). Validity in content analysis. In Mochmann

(Ed.), Computerstrategien die kommunikationsanalyse (pp. 69-112). Frankfurt, Germany: Campus. Retrieved from http://repository.upenn.edu/asc_papers/291

24.

Lee

S. Y.

Carroll

C. E.

(2011). The emergence, variation and evolution of CSR on the media and public agenda. Journal of Business Ethics, 104 (1), 115-131. doi:10.1007/s10551-011-0893-y

25.

Lombard

Snyder-Duch

Bracken

C. C.

(2002). Content analysis in mass communication: Assessment and reporting of intercoder reliability. Human Communication Research, 28, 587-604. doi:10.1111/j.1468-2958.2002.tb00826.x

26.

Lowe

Benoit

Mikhaylov

Laver

(2011). Scaling policy positions from coded units of political texts. Legislative Studies Quarterly, 36, 123-155. doi:10.1111/j.1939-9162.2010.00006.x

27.

McCombs

M. E.

(1997). Building consensus: The news media’s agenda-setting roles. Political Communication, 14, 433-443. doi:10.1080/105846097199236

28.

McCombs

M. E.

(2005). A look at agenda-setting: Past, present and future. Journalism Studies, 6, 543-557. doi:10.1080/14616700500250438

29.

McCombs

M. E.

Ghanem

. (2001). The convergence of agenda setting and framing. In Reese

Gandy

Grant

(Eds.), Framing public life (pp. 67-81). Mahwah, NJ: Lawrence Erlbaum.

30.

McCombs

M. E.

Llamas

J. P.

Lopez-Excobar

Rey

(1998). Candidate’s images in Spanish elections: Second level agenda-setting effects. Journalism & Mass Communication Quarterly, 74, 703-717. doi:10.1177/107769909707400404

31.

McCombs

M. E.

Reynolds

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

Oliver

M. B.

(Eds.), Media effects, advances in theory and research (pp. 1-16). New York, NY: Lawrence Erlbaum.

32.

McCombs

M. E.

Shaw

D. L.

(1972). The agenda setting function of the mass media. Public Opinion Quarterly, 36, 176-187.

33.

Meijer

(2004). Does success breed success? Effects of news and advertising on corporate reputation. Amsterdam, The Netherlands: Aksant Academic.

34.

Meijer

Kleinnijenhuis

(2006). News and corporate reputation: Empirical findings from the Netherlands. Public Relations Review, 32, 341-348. doi:10.1016/j.pubrev.2006.08.002

35.

Mizerski

R. W.

(1982). An attribution explanation of the disproportionate influence of unfavorable information. Journal of Consumer Research, 9, 301-310.

36.

Neuendorf

K. M.

(2002). The content analysis guidebook. Thousand Oaks, CA: Sage.

37.

Pew Research Center. (2012, June 13). 72% of Americans follow local news closely. Retrieved from http://pewresearch.org/pubs/2238/local-news-enthusiasts-newspaper-television-internet-communities

38.

Pollock

J. C.

(1995). Comparing city characteristics and newspaper coverage of NAFTA. Mass Communication Review, 22, 166-177.

39.

Pollock

T. G.

Rindova

V. P.

(2003). Media legitimation effects in the market for initial public offerings. Academy of Management Journal, 46, 631-642. doi:10.5465/30040654

40.

Ponzi

L. J.

Fombrun

C. J.

Gardberg

N. A.

(2011). RepTrak ™ pulse: Conceptualizing and validating a short-form measure of corporate reputation. Corporate Reputation Review, 14, 15-35. doi:10.1057/crr.2011.5

41.

Ragas

M. W.

(2010). Agenda-building and agenda-setting in corporate proxy contests: Exploring influence among public relations efforts, financial media coverage and investor opinion (Unpublished doctoral dissertation), University of Florida, Gainesville.

42.

Reese

S. D.

Danielian

L. H

. (1989). Intermedia influence and drug issue: Converging on cocaine. In Shoemaker

P. J.

(Ed.), Communication campaigns about drugs: Government, media, and the public (pp. 29-45). Hillsdale, NJ: Lawrence Erlbaum.

43.

Reputation Institute. (2010). 2010 Forbes/reputation institute U.S. RepTrak™ Pulse study. Retrieved from http://www.reputationinstitute.com/

44.

Reputation Institute. (2012). 2012 Forbes/reputation institute U.S. RepTrak™ Pulse study. Retrieved from http://www.reputationinstitute.com/

45.

Richey

M. R.

Koenigs

R. J.

Richey

H. W.

Fortin

(1975). Negative salience in impressions of character: Effects of unequal proportions of positive and negative Information. Journal of Social Psychology, 97, 233-241. doi:10.1080/00224545.1975.9923343

46.

Sample Size Calculator. (n.d.). Retrieved from http://www.surveysystem.com/sscalc.htm

47.

Schudson

(1984). Advertising, the uneasy persuasion: Its dubious impact on American society. New York, NY: Basic Books.

48.

Severin

Tankard

(2001). Communication theories: Origins, methods, and uses in the mass media (5th ed.). New York, NY: Longman.

49.

Shaw

D. A.

Sparrow

B. H.

(1999). From the inner ring out: News congruence, cue-taking and campaign coverage. Political Research Quarterly, 52, 323-351. doi:10.1177/106591299905200204

50.

Sheiham

Marmot

Rawson

Ruck

. (1987). Food values: Health and diet. In Jowell

Witherspoon

Brook

(Eds.), British social attitudes: The 1987 report (pp. 92-119). London, England: Cower.

51.

The Alliance for Audited Media. (2013, April 30). Top 25 U.S. newspapers for March 2013. Retrieved from http://auditedmedia.com/news/blog/top-25-us-newspapers-for-march-2013.aspx

52.

The Harris Poll. (2013, February). The Harris Poll 2013 RQ® summary report. Retrieved from https://www.harrisinteractive.com/vault/2013%20RQ%20Summary%20Report%20FINAL.pdf

53.

Tunstall

. (1996). Newspaper power: The new national press in Britain. Oxford, UK: Clarendon Press.

54.

Walker

(2010). A systematic review of the corporate reputation literature: Definition, measurement, and theory. Corporate Reputation Review, 12, 357-387. doi:10.1057/crr.2009.26

55.

Wartick

S. L.

(1992). The relationship between intense media exposure and change in corporate reputation. Business & Society, 31, 33-42. doi:10.1177/000765039203100104

56.

Wartick

S. L.

(2002). Measuring corporate reputation. Business & Society, 14, 371-392. doi:10.1177/0007650302238774

57.

Weaver

D. A.

Bimber

. (2008). Finding news stories: A comparison of searches using LexisNexis and Google News. Journalism & Mass Communication Quarterly, 85, 517-532.

58.

Wooldridge

J. M

. (2003). Introductory econometrics: A modern approach. Cincinnati, OH: South-Western College Publication. doi:10.1177/107769900808500303

59.

Wry

Deephouse

McNamara

(2006). Substantive and evaluative media reputations among and within cognitive strategic groups. Corporate Reputation Review, 9, 225-242. doi:10.1057/palgrave.crr.1550031

60.

Zhang

(2014). The repeat-Purchase contract enforcement role of business news. Corporate Reputation Review, 17, 171-182. doi:10.1057/crr.2014.7

61.

Zyglidopoulos

S. C.

Gerorgiadis

A. P

. (2006, October). Media visibility as a driver of corporate social performance. Working paper of Judge Business School, University of Cambridge. Retrieved from https://www.jbs.cam.ac.uk/fileadmin/user_upload/research/workingpapers/wp0616.pdf