Lecture 1 Flashcards
What is content analysis?
A research method to analyse characteristics of registered content of communication (written, verbal and visual)
Quantitative analysis of content characteristics of media messages
Analysing is systematic (pre agreed rules), objective and replicable
Content characteristics: manifest to latent, but always measurable and quantifiable
What is CA?
Huge. Visibility, bias, dangerous, news frames, health news, social media, movie series, content for children
WHY content analysis?
Content analysis fundamental to theory building in Communication Science.
Content is central to processes of communication
Cause of - and precondition for - media effects
Outcome of characteristics - and surrounding - media producers
> Makes possible testing relationships throughout the entire process of communication
Centrality model:
- Antecedent conditions
- Media content
- Media effects
WHY (1)
Content analysis is crucial to any theory dealing with impact or antecedents of content
In the long run one cannot study mass communication without studying content
Absent knowledge of the relevant content, all questions about the processes generating the content or the effect that content produces - are meaningless.
WHY (2)
Content analysis can help strengthen designs of media effects research:
Data from content can be linked to data from survey to explain media effects
Relevant media content is the IV in media impact
As a measure (of opportunity to) exposure, content is superior to respondent self-report.
Measuring content = explicating media impact,
Not measuring: speculation.
Selected analysed messages can serve as realistic (representative) stimulus material in experiment
Findings will have higher ecological validity
WHY (3)
Offers insights and raises questions about our mediated reality
How accurate, complete, realistic is media content?
> Media performance (media bias) (normative)
How does media content originate? Whose reality? (outcome, and who owns the me
> Power relations - those with power can control better than those with little power.
WHY (4)
Content analysis has clear practical advantages:
Accessible:
Easy access to media and messages
Low cost
Media messages always consent
UNOBTRUSIVE -> nonreactive (not interfering anything - in survey we do)
Media messages never react (we are not bothering them)
Longitudinal in no time - retrospectively a time machine (go back decades, archives).
Doing CA (with human coders)
RQ and hypo
Design measures, code book etc
Train coders (refine codebook)
Pre-test reliability of coding
Repeat steps 3-4
Draw samples and reliability tests
Code sampled media messages
Formally test reliability of coding
Analyse study data
Report study
Research questions
Descriptive RQ: questions about media content only. In isolation of its production (causes) or consumption (effects)
Answers describe (variation in) media content in some period
Value of description in scientific content analysis
Media content may be assessed against some normative standard
Essential as 1st (explorative) phase of broader research program
- Normative standards:
Descriptive RQ: example: Are Whites overrepresented as victims?
Three steps process (media content)
Explore new research territory by analysing meaning or potential causes and or effects of analysed media content
Make reasoned inferences from results about meaning or potential causes and or effects of analysed media content
Test (causal) relationships with media content as cause or outcome in future research
Explanatory research questions:
Questions linking media content to suspected cause or consequence of that content
Content posited as explanation (cause) or as itself being explainable (outcome)
Content-as-cause: to what extent does variation in media explain variation in media consumer reactions to this content?
Content-as-outcome: to what extent is variation in media itself explained by variation in characteristics of our surrounding - the producers of this content?
Media content as outcome or cause
Media producer characteristics → media content characteristics → media consumer characteristics
Comparative questions
Compare different populations of messages of media content
Answers describe differences in content between the message populations compared
Comparing media categories
Media worker - micro level: categories an individual characteristics of media workers directly responsible for producing media (male female, white non white etc)
Media organisation categories - meso level: categories of characteristic of the company in which media workers operate (public private, left liberal or conservative)
Media environment categories - macro level: categories of characteristics of the broader geographical environment in which media organisations operate.
((macro-level comparisons are often cross-national). (individualist collectivist, high vs low media regulation))
Time categories
Before and after an intervention
Historically distinct eras
Comparing theme categories
Theme related events (what was the framing - terrorism or act of violence)
Context different but type of event same
Value of comparison in scientific content analysis
Allows for testing relationship between a message population variable and message content variable
Essential for theory building concerning effects on media content
Often cannot isolate true causes of observed differences in content but does allude to hint at these causes
Facilitates making reasoned inferences about causes of content: why these observed differences between these message populations?
Description: less theoretical, less knowledge gain, less scientific (has real value)
Explanation: more theoretical, more knowledge gain, more scientific
Population (universe): all units about which a CA wants to make a claim
Census: All units in the population are analysed
Often impossible
Unnecessary if sample is representative
Probability samples: purely by chance, each member equal chance (full overview of population → sampling frame)
Sampling frame: list with all numbered units in population available for selection
Prob sample: in principle representative
Representative: a relevant characteristic of media content in the sample distributed in same way as population
Pre-condition: for generalising from sample to population (external validity)
Non-prob sampling - samples created by choice - researchers chooses units
Convenience sample:
Chosen messages were directly available
Choices based on practical considerations
Purposive sample:
Chosen messages were considered meaningful to the study
Choices based on substantive considerations
(like a magazine or something)
Consecutive unit sample:
> Suitable if analysing continuing event or story in specific period - makes possible analysing evolution of story
Not suitable for longer periods
Risk: missing large variation in content over time
Especially with news: unpredictable and diverse
All relevant messages published in a chosen time period
Time period: uninterrupted series of time units
Common version of time purposive sample
Simple random sampling: chance based sampling of n units from sampling frame
Not always effective:
No control over dispersion spread of selected units across sampling frame
Risk: selection bias (under or over representation across time)
Problematic for small samples, from long periods
Sample from long period needs to be large to be representative
Systematic sampling:
Drawing every unit from sampling frame following randomly determined starting point
N is skip interval, depends on desired sample size
Main advantage: control over dispersion over units selected from sampling frame
Even dispersion (across time)
Highly recommended for small samples, samples from long periods
Stratified sampling
Subpopulations based on characteristics meaningful
Often: stratify units by some media category (source) or time unit
Often: share of each stratum in sample proportion to size of stratum in population
Recommend: population that is diverse.
Ensure sample reflects full range of content present in population across media landscape
Stratify by media category:
List of all songs on top 10 over 19 months
Stratified according to genre: pop, hiphop, etc (sampling frames)
Each genre equally: simple random sample of 50 songs from each genre’s sampling frame
2 common sampling problems to overcome:
All prob techniques require a sampling frame.
The units to sample are not identical to unit in analysis
Cause: Lots of content is offered in clusters.
2 dimensions of clustering:
Spatial dimension: media channels or productions
Television news - cluster of news stories.
Netflix - cluster of episodes. Influencer - cluster of posts.
Temporal dimension: time units
Edition (publication date): custer of news stories
Season: cluster of episodes
Units of analysis:
About which the study makes a claim (RQ)
Sampling units: units randomly selected per step of the sampling process for inclusion in the sample: multiple steps=multiple types of sampling units.
Sampling units: 1. Titles, 2. Publication dates, 3. Stories about the UK general election
Multistage sample
Sample created through multiple rounds of sampling
First stages: stratified or cluster along spatial and temporal dimensions of media content
Recommended if sample should reflect multiple dimensions of content
Titles often not random but purposive
Editions often random using constructed period sampling
Constructed period sampling: probability sampling of time units
Ensures equal representation of selected time units across research
All time units (days of week) are even
Recommended for analysis of periodical media over long periods
Periodical media: regular intervals
WHY
The rhythm and logic of periodical media cause systematic and predictable variation repeating patterns in content: cyclical variation
If ignored: selection bias
Constructed period sampling eliminates such bias
Effective and efficient: representative but not large
Systematic version:
Skip interval chosen in such way that all publication are represented
7 days + 1 day to create 3 constructed weeks over 6 months
Even temporal dispersion of each publication
3 constructed weeks: each day of week selected 3 times - equal representation
Stratified version
Entire period divided into time strata: for each stratum: a sampling frame from which to sample
Example: time strata; all days of the week → 7 sampling frames
2 constructed weeks: random two mondays, two tuesdays etc
Nexis uni and probability sampling:
Advantages: many relevant articles quickly
Easy to create a sampling frame listing all relevant articles from random sample (clusters are circumvented)
Disadvantages: Search phrases have flaws
Subjective choice by researcher
Possibly not optimised enough of identifying all and only relevant articles
Completeness of database is questionable
Some key titles not available, relevant articles not available, visual content not archived.
Nexis uni and prob sample
Does a sampling frame offer a full overview?
Does each article from the population have an equal chance? Questionable.
Certainty not…
