• Aucun résultat trouvé

Hypothesis 1 – Centrist newspapers have a higher bias

Dans le document Media bias and media firm strategy (Page 120-0)

3. Media bias, INSTITUTIONAL environment AND new and Sensitive technologies

4.5 Resource-based analysis: flexibility versus commitment

4.6.1 Hypothesis 1 – Centrist newspapers have a higher bias

As argued by Mullainathan and Schleifer (2005), politically biased newspapers may already have their job cut out, in the sense that they focus on biasing the political stories more either to the left or right as per their political tilt. This has also been shown by numerous other authors as indicated earlier in this paper. Thus, centrist papers need to find interesting stories which they may wish to bias according to consumer perceptions of a specific story.

It has been shown very clearly by Eurobarometer data that European consumers, as a majority, express a negative attitude towards GM foods, and thus it is certainly interesting for centrist newspapers to show more negative bias on news pertaining to GM foods as compared to the conservative and liberal newspapers. These head-line grabbing stories are likely to help centrist newspapers obtain more attention from the consumers and thus result in an increased market share of circulation. This basic premise follows the logic that by deciding to bias a story more than the conservative or liberal newspapers (accuracy), centrist newspapers believe that they can retain (or even increase) their credibility amongst their readers.

H1: Centrist newspapers will display a higher level of negative bias than conservative or liberal newspapers, ceteris paribus.

120 4.6.2 Hypothesis 2 – Newspaper market shares play a role in the bias

Newspapers in general are keen to maintain or increase their circulation market shares, and this influences the type of articles/news that appears in a particular newspaper. Thus, they search for 'hot issues' in order to differentiate themselves and retain reader interest.

Two specific research publications were identified concerning the coverage of GM foods by European newspapers. Fitzgerald, Campbell and Sivak (2002) show that while global media sentiment was very positive in 1995, since then, media coverage on GM food had moved to a neutral middle ground by the time their paper was completed in 2002. By examining press coverage of GM foods in Spain and the UK between 1999 and 2004, Vilella-Vila and Costa-Font (2008) argue that differences in media reporting along with attitudes towards journalism correlate with attitudes and risk perception, thus reinforcing the hypothesis of a media bias in newly created technology risks.

It is hence hypothesised that news on GM foods may well be one such ‘hot story’ that newspapers may wish to use when they observe a dip in their market shares, in order to differentiate themselves from other competing newspapers and thus aim to obtain increased reader attention and loyalty. Since the Eurobarometer scores indicate that Europeans are largely negative to GM foods, newspapers will have an inclination to use this negativity in order to attract their readers. Thus, according to the differentiation theory of Porter, the negative bias levels of newspapers should contribute to an increase in market shares.

However, when newspaper organizations observe that their market shares are increasing, they will be tempted to use the negative bias as a tool even more frequently. Thus, market share may be an endogenous variable since it is likely that the bias level causes an increase in circulation and when the newspaper ‘learns’ that a change in bias causes a change in market share, it may use bias as a strategic input when it observes such a change in market share. But, market share was used in this analysis as an independent variable for the want of a better alternative variable or an appropriate instrumental variable.

H2: An increase in the market shares will lead to an increase in the negative bias in these newspapers.

121 4.6.3 Hypothesis 3 – Centrist newspapers are followers

As is well known, the institutional theory argues that when placed in an ‘uncertain’

environment, organizations tend to follow each other from a strategic point of view. In our case, the ‘uncertainty’ will relate to possible reader reactions on negatively biased stories on GM foods. While the Eurobarometer scores are in favor of a negative consumer bias towards GM foods, newspapers can never be certain of consumer reactions to negatively biased stories on GM foods.

While all the newspapers will aim to reduce this uncertainty by following each other, as mentioned earlier, politically-biased newspapers will be less concerned by this phenomenon since they tend to focus on biasing political stories according to their specific political leanings.

On the other hand, centrist newspapers do not have a fixed political agenda. Thus, in order to ensure that their negatively biased stories on GM foods obtain sufficient attention, it is hypothesised that the centrist newspapers are likely to follow the conservative and liberal newspapers, as per Porter’s differentiation theory. The flexible resource approach of the centrist newspapers provides them the space to move the way they wish to (both in the extent and in the direction of the bias or in other words, both in terms of accuracy and ideology) without any major threat to their credibility amongst their readers.

H3: Centrist newspapers follow the strategies of conservative and liberal newspapers while publishing negatively biased articles on GM foods.

122 4.6.4. Hypothesis 4 – Centrist newspapers show a higher bias

This hypothesis sums up the core argument of this paper regarding resource flexibility. As explained earlier, and according the commitment versus flexibility theory advanced by Ghemawat and del Sol (1998), it is proposed that centrist newspapers have a higher flexibility in the direction (ideology) and extent (accuracy) of their bias against both political and non-political newspapers. Their flexible ideological/accuracy positioning and thus the lack of this particular firm-specific resource may ensure that they do not lose their credibility amongst their readers by the change in their type and level of bias on both political and non-political stories.

H4: Centrist newspapers will show a higher bias as compared to both conservative and liberal newspapers in both political and non-political news and their type of bias can be on either side of the bias spectrum.

4.7 Data and methodology

A total of 46 publications from 17 countries (including 4 from India as control newspapers) were retained for evaluation using content analysis after a careful selection process which is explained in Chapter 2. 3991 articles and editorials concerning genetically modified foods which appeared between the period 2005 and 2010 (6 years) were collected during that stage.

However, another 81 articles from Le Figaro were made available for this stage of the analysis which increased the total number to 4072 for the current paper. A list of 'search' words was finalised based on Fitzgerald et al. (2002). Thus, the chosen key words were:

 GMO

 Genetically Modified Food

 Genetically Engineered Food

 Genet* Modi* Foods

 Biotech food

 Gene Tech food

In order to be able to search in the local language publications in Continental Europe, the above terms were first translated into all the relevant languages.

123 Then, each issue of the selected newspapers was searched using the above key words in the respective languages and each identified article was named using a specific coding system so that its source could be identified easily at a later stage.

Articles, if not already in English, were translated into English using the Google translation software. While the quality of the translations was far from being perfect, several factors were in favor of adopting this procedure as explained in Chapter 2 of this thesis.

Bias levels in these articles were evaluated by two external analysts based on a well-established methodology in content analysis studies and is explained in Chapter 2. It was possible to prove that a negative bias against genetically modified foods does exist in European newspapers. A set of institutional factors which may affect this bias were also identified.

Data collected regarding the negative bias by the 46 newspapers during 2005 and 2010 was used in this research. In addition, articles based on the theme of ‘Christine Lagarde’ (the finance minister during 2010) and ‘tax increase’ were collected as ‘political’ news items from 3 newspapers in France – Le Figaro (conservative), Le Monde (liberal) and Le Parisien (centrist) for the year 2010.

4.8 Results

The average negative bias scores by newspaper by year for selected countries were first plotted as simple line graphs (shown as Figure 4.3).

124 Figure 4.3. Average Negative Bias by Newspaper and by Country

0 10 20 30 40 50 60 70 80 90 100

2005 2006 2007 2008 2009 2010

% Average Negative Bias by Year and by Newspaper - Germany

Bild Die Welt Die Zeit R. Post S. Zeitung

0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0

2005 2006 2007 2008 2009 2010

% Average Negative Bias by Year and by Newspaper - Austria

Der Standard Diepresse

125

% Average Negative Bias by Newspaper - France 2005-2010

% Average Negative Bias by Year and by Newspaper - UK

% Average Negative Bias by Year and by Newspaper - Italy

Stampa Cor.del.Sera Repubblica

126

% Average Negative Bias by Year and by Newspaper - Spain

% Average Negative Bias by Year and by Newspaper - Sweden

% Average Negative Bias by Year and by Newspaper - Russia

Kommersant Pravda Ross.Gazeta

127 By examining the average annual negative bias in each country, it can already be noted that the publications have more or less followed each other. Figure 4.4 provides the % of negatively biased articles by newspaper.

Insert Figure 4.4. here from the end of the chapter

__________________________________________________________________________

Table 4.2 gives the mean and standard deviation of negative bias scores by country and by year, while Table 4.3 shows the overall mean and standard deviation of the negative scores by each country.

0 10 20 30 40 50 60 70 80 90 100

2005 2006 2007 2008 2009 2010

% Average Negative Bias by Year and by Newspaper - India

Hindu TOI Telegraph

128 Table 4.2. Average Negative Bias Score by Country - Mean and SD by Year

Germany (5 newspapers)

129 Sweden (3)

Year Mean Std. deviation

2005 75.0 36.3

2006 55.6 31.4

2007 65.1 14.4

2008 47.2 21.0

2009 72.2 25.5

2010 61.6 14.4

Russia (3)

Year Mean Std. deviation

2005 64.4 3.8

2006 76.7 25.2

2007 90.6 9.1

2008 85.2 17.0

2009 73.6 20.6

2010 71.5 18.8

India (3)

Year Mean Std. deviation

2005 73.5 37.4

2006 70.4 6.4

2007 77.9 20.7

2008 60.6 9.2

2009 65.0 18.3

2010 62.9 11.2

130 Table 4.3. Average Negative Bias Score by Country - Overall Mean and SD

india 17 68.07757 15.97295 45.83333 100 Variable Obs Mean Std. Dev. Min Max . summarize india

russia 18 77.00865 17.05225 50 100 Variable Obs Mean Std. Dev. Min Max . summarize russia

sweden 17 65.82633 22.89105 25 100 Variable Obs Mean Std. Dev. Min Max . summarize sweden

uk 29 52.10067 19.20804 20 100 Variable Obs Mean Std. Dev. Min Max . summarize uk

italy 17 61.18249 18.79009 25 100 Variable Obs Mean Std. Dev. Min Max . summarize italy

spain 12 46.97014 19.70628 20 85.71429 Variable Obs Mean Std. Dev. Min Max . summarize spain

france 24 74.64567 10.56208 53.125 100 Variable Obs Mean Std. Dev. Min Max . summarize france

austria 12 63.85475 14.02035 42.85714 78.57143 Variable Obs Mean Std. Dev. Min Max . summarize austria

germany 29 63.43424 21.00601 0 100 Variable Obs Mean Std. Dev. Min Max . summarize germany

131 It can be seen from Table 4.3 that France shows the highest mean negative score as well as the lowest standard deviation amongst all the major countries studied. It can be argued that France is the country where there is a competition amongst leading newspapers and hence instead of differentiating on this particular issue of GM foods as can be expected from Porter’s competitive theory, newspapers may be tending towards the 'spin' theory proposed by Mullainathan and Schleifer (2005).

This gives an indication that it can be what Mullainathan and Schleifer (2005) call as the 'spin' bias (as compared to 'ideology' bias) where all players go in the same direction and make use of the situation to get consumers on their side by exaggerating the story in the same direction as the other players. It may also be interpreted as supporting the institutional/isomorphism effect. While discussing isomorphism, Srikantia and Bilimoria (1997) argue that subservience to the institutional models implicit in the conduct of mainstream business organization traps the discipline in a self-perpetuating cycle of normative and mimetic influences, leading the practitioners to mutually reinforce and sustain institutional forms compatible with their jointly held belief systems.

Spin bias can also be seen as a specific instance of isomorphism as follows: newspapers when facing severe competition, wish to take advantage of the issue at hand (and in this case, the consumers’ negative perception of GM foods), and thus race each other to grab reader attention by publishing negatively biased articles on GM foods. It is also possible to distinguish between isomorphism and spin bias by saying that isomorphism is present when a firm blindly follows the industry norm since it is not sure how the market will react, whereas in this case, the newspapers know that it is good to outshout each other rather than just blindly follow the others. On the other hand, one can also argue that the possible reaction of the readers may be an unknown for the newspapers.

132 Tables 4.4 to 4.6 give the mean and standard deviation by year of the annual average negative bias along with the overall mean and standard deviation for each of the 3 groups of newspapers – liberal, conservative and centrist newspapers. Table 4.7 provides the results of the paired t-tests (with unequal variances) which compare the means of the annual average negative bias of the centrist versus liberal newspapers, while table 4.8 provides the results of a similar comparison between centrist and conservative newspapers.

Table 4.4. Mean and SD – Liberal Newspapers

% Average Negative Bias by Newspaper

-Liberal Newspapers 2005-2010 StandardZeit

133 Table 4.5. Mean and SD – Conservative Newspapers

summarize conservmeans

Variable | Obs Mean Std. Dev. Min Max ---+--- conservmeans | 70 59.49448 23.28502 0 100

Conservative Newspapers

Mean and Standard Deviation of % Average Negative Bias by Year (no. of newspapers = 12)

Year Mean Std. deviation

2005 52.8 32.2

2006 64.2 22.0

2007 66.4 23.6

2008 61.4 19.7

2009 65.1 18.0

2010 63.0 22.0

0 10 20 30 40 50 60 70 80 90 100

2005 2006 2007 2008 2009 2010

% Average Negative Bias by Newspaper -Conservative Newspapers 2005-2010

Die Presse Bild Welt R.Post ABC Figaro Irish Ind.

TOI

Sv. Dagbladet Telegraph Daily Mail Times

134 Table 4.6. Mean and SD – Centrist Newspapers

summarize centristmeans

Variable | Obs Mean Std. Dev. Min Max ---+---

centristmeans | 51 69.84891 17.16629 25 10 Centrist Newspapers

Mean and Standard Deviation of % Average Negative Bias by Year (no. of newspapers = 9)

Year Mean Std. deviation

2005 68.3 13.7

2006 73.1 17.7

2007 79.1 15.7

2008 56.7 21.0

2009 75.3 16.2

2010 66.1 10.4

0 10 20 30 40 50 60 70 80 90 100

2005 2006 2007 2008 2009 2010

% Average Negative Bias by Newspaper -Centrist Newspapers 2005-2010

S. Zeitung Postimees Le Parisien Telegraph La Stampa Cor.della Sera Il Sole 24 Ore Adevarul Kommersant

135 Table 4.7. Means of Average Negative Bias – Centrist versus Liberal Newspapers X = average annual means for Centrist newspapers

Y = average annual means for Liberal newspapers

Table 4.8. Means of Average Negative Bias – Centrist versus Conservative Newspapers

X = average annual means for Centrist newspapers Y = average annual means for Conservative newspapers

It can be observed that the mean of all annual observations of the negative bias score of centrist newspapers is the highest (as compared to the liberal or the conservative newspapers). As can be seen in Tables 4.7 and 4.8, the null hypothesis that centrist papers' average annual negative scores on GM food articles are the SAME as those of the liberal papers can be more or less rejected since Pr(T>t) = 0.1014. On the other hand, the null hypothesis that centrist papers' average annual scores on GM food articles are the same as those of conservative papers can be FIRMLY rejected since Pr(T>t) = 0.0028.

Pr(T < t) = 0.8986 Pr(|T| > |t|) = 0.2029 Pr(T > t) = 0.1014

. ttesti 51 69.85 17.17 89 65.82 19.15, level (95) unequal

Pr(T < t) = 0.9972 Pr(|T| > |t|) = 0.0057 Pr(T > t) = 0.0028

. ttesti 51 69.85 17.17 70 59.49 23.29, level (95) unequal

136 The next step was to examine possible relationships amongst the variables through the Stata programme. The summary of the first set of results is given in Table 4.9. The dependent variable used in this regression was dcscore, a dummy that is 1 if the content score of the article is negative (article is negatively biased) and 0 if it is positive or neutral.

The 4 independent variables used were:

 market share of the newspaper in the country

 dtiltR = dummy for conservative tilt of the newspaper

 dtiltL = dummy for liberal tilt of the newspaper

 dtiltC = dummy for a centrist newspaper

The last 3 variables were categorised using www.worldpress.org except in the case of a few newspapers which were not included in the above website.

Table 4.9. Newspaper tilt and the negative bias of content score Dependent variable: dummy of negative bias of content score (dcsore)

Variables dcscore

Note: standard errors shown in brackets

*, ** and *** represent significant levels at 10%, 5% and 1% respectively

The coefficient of the variable market share is not significant probably because it represents an annual market share for each newspaper which is unlikely to be used as a decision variable by any newspaper in the short-term.

The significant coefficient of the centrist newspapers already indicates that the highest bias would be in the centrist newspapers, and moving from dtiltc=0 to dtiltc=1 is associated with a 7.5% increase in the negative score as seen in the marginal effects, with a standard deviation of 0.019.

137 But this certainly deserves more investigation. In particular, it would be interesting to know if liberal newspapers tend to bias more in countries with a conservative majority. However, as explained later on, the limited database in this current analysis does not provide such an opportunity, and this is left for future research. A multinomial logit stata programme was also used with cscore as the dependent variable and the same 4 independent variables as above, and the results reconfirm the results from the probit analysis.

The above results seem to suggest that newspapers are indeed trying to use media bias to create some differentiation, but this appears to be applicable only in the case of journals that are relatively 'undifferentiated' (from a political point of view) in the first place. Thus, in our case, the centrist papers appear to have a distinct advantage since they have the possibility to bias in any direction and to any extent and still retain their credibility. The newspapers which are already politically differentiated appear to be somewhat restrained in their bias on political and non-political stories probably due to the inherent fear of losing credibility amongst their readers. In addition, they do not change their ideological positioning.

However, yearly market share data is not completely satisfactory as a variable since it is not very useful while taking short-term decisions. This was also evident when even a lagged (by a year) market share variable failed to give any satisfactory results as shown in Table 4.10.

Table 4.10. Probit Regression Results - Centrist papers but with one year lag for newspaper market shares

Dependent variable: dummy of negative bias of content score (dcsore)

Variables dcscore Market share with a year lag

Dummy for centrist

Note: standard errors shown in brackets

*, ** and *** represent significant levels at 10%, 5% and 1% respectively

138 An OLS regression was carried out with the average annual negative bias by newspaper as the dependent variable and the same two independent variables - annual market share and the dummy centrist newspaper. It was also decided to test the possible interaction between the two independent variables. In this case, only 185 entries were used since the average bias for n number of articles in a particular year was considered as a single (and not a multiple) entry. Similar results were obtained as shown in Table 4.11.

Table 4.11. Results of OLS Regression with interaction Variables Note: standard errors shown in brackets

*, ** and *** represent significant levels at 10%, 5% and 1% respectively a) Regression using ‘average negative bias by year by newspaper’ as

dependent variable and ‘annual market share by newspaper’ and ‘centrist newspaper dummy’ as the independent variables with an interaction between the two independent variables

b) Regression using ‘total number of articles on GM foods by year by

newspaper’ as dependent variable and ‘annual market share by newspaper’

and ‘centrist newspaper dummy’ as the independent variables with an interaction between the two independent variables

The results show that newspapers tend to increase the number of articles on GM foods as they see their annual market share decrease during the year and this may then lead to an increase in their market share. Thus hypothesis 1 cannot be rejected. However, in both cases, the interaction variable was not significant.

139 Hence, it was decided to focus on monthly circulation/market share data of newspapers.

After making significant efforts to identify detailed monthly market share data for all the 47

After making significant efforts to identify detailed monthly market share data for all the 47

Dans le document Media bias and media firm strategy (Page 120-0)