Uppsala University Department of Statistics Bachelor Thesis, Spring 2017 Supervisor: Mattias Nordin Mass media’s influence on attitudes towards the EU Do people with different levels of news consumption differ in their attitude towards the EU? Madeleine Larsson
35
Embed
Mass media’s influence on attitudes towards the EUuu.diva-portal.org › smash › get › diva2:1107916 › FULLTEXT01.pdf · EUupplysningen (2016) showed that around 30% of all
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Uppsala University Department of Statistics Bachelor Thesis, Spring 2017 Supervisor: Mattias Nordin
Mass media’s influence on attitudes
towards the EU Do people with different levels of news consumption differ in their
attitude towards the EU?
Madeleine Larsson
Abstract
The news media is an important institution for all democracies. It helps the citizens to
keep informed and be able to take part of the public debate, but in recent years the
gap between the active and the inactive news consumer has increased. Does it
make any difference? In order to contribute to the field, this research paper is to
make a quantitative analysis to look at whether people with a high consumption of
news from the Swedish mass media differ in their attitude towards the EU.
As an ordered logistic regression was not applicable when analyzing the categorical
dependent variable, that are measuring attitudes towards the EU, three binary
logistic regressions was instead used. The results show that individuals with a high
consumption of news from the Swedish mass media have higher odds of having an
opinion of a positive attitude toward the EU. The data used are however self
provided and voluntary surveydata, which contain various biases. The fact that it is
only observed and not experimental data makes it impossible to estimate a causal
effect, which instead is up to future research.
Keywords:
Attitudes, EU, news consumption, logistic regression, Swedish mass media
1
Table of contents
Introduction 3
1. Theory 4 1.1 Background 4
2. Data 6 2.1 Swedish mass media 8 2.2 Descriptive statistics of Attitudes towards EU 9
Table 1 Generally speaking what is your attitude towards the EU? 10
3. Methodology 10 3.1 The model 11
3.1.1 Probabilities 11 3.1.2 Odds 12
3.2 Independent variables 13 Table 2 The Dependent Variable 15 Table 3 Personal Attributes 15 Table 4 Professional life 15 Table 5 European background 16 Table 6 Political belief 16 Table 7 News Consumption 17
3.3 Levels of media consumption 17 3.4 The “no opinion” category 18
4. Results 19 4.1 “No opinion” vs. “opinion” 19
Table 8 “No opinion” vs. “Opinion” 20 4.2 Analysis of attitudes towards the EU 21
4.2.1 Positive Attitude vs. Not Positive Attitude 22 Table 9 Positive attitude vs. Not Positive Attitude 22
4.2.2 Not Negative vs. Negative Attitude 22 Table 10 Not Negative attitude vs. Negative Attitude 23
4.2.3 Neutral attitude vs. Not Neutral Attitude (Positive + Negative) 23 Table 11 Not Neutral attitude vs. Neutral Attitude 25
* basic model = Big Frequency & Quantity of news consumption ** Personal attributes = sex and age *** Professional life =education and income of the household **** European background = Childhood country (within EU) for you, your parents or both*****variable interestinp.
The dependent variables has reference group 0 in the model (people with no
opinion) meaning that the odds of having an opinion increases as an individual
consume news both more frequently and/or from more sources. This is an interesting
finding, which can make the next analysis questionable as it can be the case that
consuming a lot of news not gives you a particular attitude but rather an attitude (at
all).
From the analysis, it looks as if people not consuming a lot of news has a higher
tendency of answering “no opinion”. How many and what makes someone choose
this option instead of the neutral category, is however still unclear. In the overall
analysis (with all five surveys) the “no opinion” category is rather small (only around
4%, see table 1) which lead to the conclusion that the selection bias created by
removing these observations probably will be smaller than combining it with the
neutral category for the next analysis. To establish that people who consume a lot of
news has a bigger tendency to have an opinion at all, is however insightful, and a
20
step closer to answer the question of how (if at all) the Media influence attitudes
towards EU.
4.2 Analysis of attitudes towards the EU As the response data is ordinal an ordered logit regression would be the most
intuitive model to use as it takes the differences and similarities between all five
categories into consideration. A binary logit model only distinguishes between two
groups. The ordered logit model was therefore first attempted, but as the
proportional odds assumption failed for the model, the inference could not be
statistically assured.
The proportional odds assumption states that the odds ratios should be the same
between all the dependent variable’s ordered categories. With other words, the slope
estimate between each pair of outcomes across two response levels should be the
same, no matter which pair studied. The assumption of proportional odds is very
strong for the ordered logit model, which also has given the method its second name
of ‘proportional odds model’ (Williams, 2016). On the other hand, by combining
observations into fewer categories less information can be extracted, as the
categories cover a bigger range of observations which therefore become less
specific. As more information can be extracted from the proportional odds model,
another attempt was tried but with fewer levels positive, neutral and negative. Also
this test failed because of the same reason as before.
Three separate binary logistic models were therefore created so that each pair of
outcomes Positive vs. Not positive (a group of negative and neutral), Negative vs.
Not Negative, Neutral vs. Not Neutral instead could be compared individually. By
doing this the odds can vary between the groups, and the proportional odds
assumption is no longer required.
21
4.2.1 Positive Attitude vs. Not Positive Attitude
To evaluate the correlation of consuming a lot of news and having a positive attitude
towards EU, a binary logit model between the positive observations and a combined
group of negative and “neither positive or negative” observations was performed. All
models proved significant, likewise did the variable of interest perform a good overall
results.. The “ Not positive attitude” is the reference group of the analysis, meaning
that the odds of having a positive attitude increase when consuming more news.
Table 9 Positive attitude vs. Not Positive Attitude
Model Tests
Basic* + Personal
attributes**
+Professional life***
+European background****
+ Prior beleifs*****
Variable Statistics
Odds Ratio
1.114 1.304 1.264 1.280 1.194
Pvalue 0.0371 <.0001 <.0001 <.0001 0.0052
95 % CI 1.006 1.234
1.163 1.462
1.125 1.420 1.139 1.439 1.054 1.351
Overall model statistics
Testing global null
:H 0 Beta=0
(rejection of)H 0
(rejection of)H 0
(rejection of )H 0
(rejection of )H 0
(rejection of)H 0
df 1 3 6 9 18
Nagelkerke R2
0.0009 0.0094 0.0596 0.0633 0.1717
No. of obs (pos. / not pos.)
3170 / 4606
3170 / 4605 2945 / 4242 2945 / 4242 2945 / 4242
* Basic model = Frequency & Quantity of news consumption, ** Personal attributes = sex and age *** Professional life =education and income of the household **** European background = Childhood country (within EU) for you, your parents or both. *****Prior beliefs = Political party affiliation & interest in politics.
4.2.2 Not Negative vs. Negative Attitude
The second test performed was between the observations with a negative attitude
towards EU versus “not a negative attitude” (= positive and neutral attitude). The
negative group is the reference group of the model. The basic* model that only
22
include consumption of news, did not prove significant but notably improved as more
background variables was added.
The overall result confirms the previous model and show that consumption of news
has a positive correlation with the attitude towards EU. With other words people
consuming more news have higher odds of not having a negative attitude. When
including all control variables, the 95 % confidence interval of the odds ratio is
between 1.02 to 1.328, meaning that people consuming a lot of news have 1.02 to
1.328 times higher odds of not to have a negative attitude.
Table 10 Not Negative attitude vs. Negative Attitude
* Basic model = Big Frequency & Quantity of news consumption ** Personal attributes = sex and age *** Professional life =education and income of the household **** European background = Childhood country (within EU) for you, your parents or both. *****Prior beliefs = Political party affiliation & interest in politics.
4.2.3 Neutral attitude vs. Not Neutral Attitude (Positive + Negative)
In the last model the odds of consuming more news while having a positive or
negative attitude is compared to the odds of consuming a lot of news while having a
23
category 3 opinion (in the scale of 5, meaning that the individual is neither positive or
negative). The test was carried out to see if people consuming more news get a
broader perspective and therefore tend to be in the middle of the scale. This can
seem similar to the first analysis between “opinion” and “no opinion”. However, the
difference being that the underlying reasons of the category 3 answers are within the
ordered scale, and is therefore an attitude, which “no opinion” cannot be proven to
be. An interpretation and comparison with only the other two levels are therefore now
possible. The results show that the basic* model (only including the media
consumption variable) is not statistically significant. As background variables are
added the overall model improves and becomes more powerful, even if there are
almost no advancement in the explanation rate. The single variable of media
consumption is only significant in the second model. This show that the model, and
more importantly news consumption is not a good variable to distinguish between
these groups. There is thus no difference in news consumption between someone
that neither are pro or against EU, and the two groups with more defined positions.
24
Table 11 Not Neutral attitude vs. Neutral Attitude
* basic model = Frequency & Quantity of news consumption ** Personal attributes = sex and age *** Professional life =education and income of the household **** European background = Childhood country (within EU) for you, your parents or both. *****Prior beliefs = Political party affiliation & interest in politics.
4.3 Overall analysis
Comparing all the results side by side the test of “opinion vs. no opinion”, a little
unexpectedly, yields the highest numbers. Consumption of news has the highest
odds ratio between these two groups and the overall model has the highest
explanation rate. This, even as fewer variables are included than in the rest of the
models. Both the test of “positive vs. not positive attitude” and “negative vs. not
negative attitude” clearly points in the same direction, and show that people
consuming more news have higher odds of having a positive attitude towards the
EU. It is however a slight difference in the odds ratios (1.19 and 1.16), where the
odds ratio is lower when positive and neutral attitude is compared to the negative.
This indicates that the neutral group have decreased the odds when added to the
positive. The logistic regression between “neutral attitude and not neutral attitude”
both provide the lowest explanation rate but more importantly has insignificant odds
25
ratios for consumption of news. The level of news consumption is therefore not
significantly different between these two constructed groups. The overall analysis
that can be drawn when comparing these tests are therefore that people consuming
more news have higher odds of simply having an opinion/attitude in the question of
EU. Why people don’t have an opinion is on the other hand unclear. Within the
attitude scale which the odds are in favor for when consuming a lot of news, the
odds are in a second stage also higher for a positive attitude (when consuming a lot
of news).
Table 12 Overall Analysis
Analysis Variable Statistics Overall model Nagelkerke R2
Opinion vs no opinion
Odds ratio
Pvalue
1.851
<.0001
0.1819
Positive vs. Not Positive
Odds ratio 1.194 0.1717
Pvalue 0.0052
Negative vs. Not Negative
Odds ratio 1.164 0.1511
Pvalue 0.0241
Neutral vs. Not Neutral
Odds ratio 1.048 0.0519
Pvalue 0.4751
* Results are from full models ** underlined group = reference group
To summarize The odds are 1.85 times higher of having an opinion when
consuming a lot of news. If an individual does have an opinion and consume a lot of
news, the odds are 1.19 times higher of that opinion to be positive towards EU (than
a neutral or negative attitude). When combining the neutral attitude observations with
the group of positive attitudes the odds ratio of news consumption decreases by
0.03, indicating that there are lower odds of being neutral than positive when the
news consumption is high.
26
5. Discussion and Conclusion
The fact that there is a differences between individual’s attitude towards EU when
consuming a lot of news can thus be verified by the logistic regression. The odds are
higher of having an opinion when news consumption is high. When having an
opinion, it is higher odds of the attitude to be positive. This confirms the hypothesis
that the paper stated in the beginning of this paper, however, correlation does not
mean causation. Do individuals develop the positive attitude because they consume
a lot of news, or do they consume a lot of news because they have an interest and
by that have an attitude/opinion?
To capture the true causation, or as for this study, the influence of mass media on the
opinion of the EU, an experimental study would be necessary. By randomizing the
consumption of news between individuals, and comparing the attitudes of EU with a
control group after the experimental period is over, the data would be of better quality
and with less bias. Both OVB, SEB, Social desirability bias and selection bias would
be controlled for. The only variable differentiating between the groups would further
be the news consumption, and therefore the only variable possible to influence the
attitude. This experiment would in practice, be extremely difficult to realize. Firstly,
the experiment would have to last for a rather long period of time to capture the
effects, and it would therefore be hard to find (randomly chosen) participants. To
avoid the news is in addition rather hard in today’s society, where it can be accessed
from various sources, which always are surrounding us, whether we like it or not.
However, the intention of the essay was never to provide all answers. Rather to
conduct a first analysis of the influence of the Swedish mass media by consumption
of news in the 21st century, for future research to build on. So to answer the
research question of “Does people with a high consumption of news differ in their
attitude towards the EU?” The thesis concludes that there is a correlation between a
positive attitude and a big consumption of news, which confirms the hypothesis in
the beginning of the paper. The causation of why the people reading more news
have a positive attitude is however still unanswered.
27
References
Aalberg, T., Blekesaune, A. & Elvestad, E.. (2013). Media Choice and Informed Democracy: Toward Increasing News Consumption Gaps in Europe?. The International Journal of Press/Politics . 18 (No. 3), pg. 281 303. Aalberg, T., Strömbäck, J. & De Vreese, C. . (2012). The framing of politics as strategy and game: A review of concepts, operationalizations and key findings. Journalism (London, England) . 13 (issue 2), pg. 162178. Antonakis, J. . (2010). On making causal claims: A review and recommendations. The Leadership quarterly . 21 (No. 6), ISSN: 10489843, DOI: 10.1016/j.leaqua.2010.10.010, pg. 10861120. Antonakis, J., Bendahan, S., Jacquart, P., & Lalive, R. (2014). Causality and endogeneity: Problems and solutions. In D.V. Day (Ed.), The Oxford Handbook of Leadership and Organizations (pp. 93117). New York: Oxford University Press. Berg, L. & Bové, K. (2016). Svenskarnas attityder till EU har stabiliserats Analys av SOMinstitutets undersökning 2015. Europapolitisk Analys. Svenska Institutet för Europapolitiska Studier. 9epa. pp. 1 12 Converse, J.M.. (1977). Predicting No Opinion in the Polls. Oxford University Press on behalf of the American Association for Public Opinion Research . 40 (No. 4), pp. 515530. EUUpplysningen. (2016). EUlagar gäller framför svenska lagar. Available: http://www.euupplysningen.se/SverigeiEU/EUlagargallerframforsvenskalagar/. Last accessed 20170427. Foltz, B.. (2015). Statistics 101: Logistic Regression Probability, Odds, and Odds Ratio . Available:www.youtube.com/watch?v=ckkiGSDuV8&index=2&list=PLIeGtxpvyGJmBQ9XoFD4rsb3hkcX7Uu. Last accessed 20170521 Foltz, B.. (2015) .Statistics 101: Logistic Regression, Logit and Regression Equation. Available:https://www.youtube.com/watch?v=NmjT1_nClzg&index=3&list=PLIeGtxpvyGJmBQ9XoFD4rsb3hkcX7Uu. Last accessed 20170521 Johansson, B. (2011). Partipropaganda i spalterna?. In: Lycksalighetens ö . Göteborg: University of Gothenburg. pp. 421434. ISBN: 9789189673212 Krosnick, J., Holbrook, A., Berent, M., Carson, R., Hanemann, M., Kopp, R., Mitchell, R., Presser, S., Ruud, P., Smith, V., Moody, W., Green, M. & Conaway, M.. (2002). THE IMPACT OF “NO OPINION” RESPONSE OPTIONS ON DATA QUALITY:
28
NONATTITUDE REDUCTION OR AN INVITATION TO SATISFICE?. Public Opinion Quarterly . 66 (Issue 3), pp. 371 403. Mendenhall, W. & Sincich, T.. (2014). . In: A second Course in statistics Regression Ananlysis . Essex, England: Pearson New International Edition. pg. 455. Maier, J. & Rittberger, B.. (2008). Shifting Europe’s Boundaries: Mass Media, Public Opinion and the Enlargement of the EU. European Union Politics . 9 (issue 2), 243–267, DOI: 10.1177/1465116508089087 Mutz, D., & Soss, J.. (1997). Reading Public Opinion: The Influence of News Coverage on Perceptions of PublicSentiment. The Public Opinion Quarterly. 61 (No. 3), pp. 431451. Nationalencyklopedin, massmedier . http://www.ne.se/uppslagsverk/encyklopedi/lång/massmedier (Last accessed: 20170509) Nico Drok & Liesbeth Hermans (2016) Is there a future for slow journalism?, Journalism Practice, 10:4, 539554, DOI: 10.1080/17512786.2015.1102604 Peduzzi, P., Concato, J., Kemper, E., Holford, T.R. & Feinstein, A.R.. (1996). A Simulation Study of the Number of Events per Variable in Logistic Regression Analysis. J Clin Epidemiol . 49 (no. 12) , pp. 1373–1379 Prat, A. & Strömberg, D.. (2011). The Political Economy of Mass Media. CEPR Discussion Paper . No. DP8246, Available at SSRN: https://ssrn.com/abstract=1763655. Statistiska Centralbyrån. (2014). VANLIGARE ATT RÖSTA I RIKSDAGSVALET PÅ SENARE ÅR. Available: http://www.scb.se/hittastatistik/sverigeisiffror/valochpartier/valdeltagande/. Last accessed 20170427. Vernersdotter, F.. (2015). KODBOK Den nationella SOMundersökningen. SOMInstitutet 2015 . Göteborgs Universitet, pp. 1. Williams, R.. (2016). Understanding and interpreting generalized ordered logit models. The Journal of Mathematical Sociology, . 40 (1), pg. 7 20. DOI: 10.1080/0022250X.2015.1112384
29
Appendix Part 1 The Questions Vernersdotter, F.. (2015). KODBOK Den nationella SOMundersökningen. SOMInstitutet 2015 . Göteborgs Universitet, pp. 13 58.
Media Consumption
30
Personal Attributes
Professional life
31
European Background
Prior beleifs
32
Part 2 SAS Statistical Output 2.1 Age variable
33
2.2 Spearman correlation Coefficients
ofta sex manga hushink oftamanga
alder utb utlandsfodd utlandsfoddf
ofta 1.00000 0.03313 0.54349 0.04284 0.49627 0.33492 0.01715 0.07132 0.06895
sex 0.03313 1.00000 0.03388 0.03746 0.03365 0.02682 0.11080 0.01480 0.02101
manga 0.54349 0.03388 1.00000 0.04312 0.68158 0.34930 0.01665 0.05735 0.06357