Female Voices in the News: Structural Conditions of Gender Representations in Norwegian Newspapers

Female Voices in the News

Research on the capacities in which women feature in the news tends to focus either on gender representations in news texts, or on organizational-, professional-, and labor-related aspects of news production. Studies either measure quantitatively the number of women quoted in news reports, analyze source selection among professional reporters, perform case study analyses of press coverage of female political candidates, or analyze gender in various journalistic genres. This line of research assumes that the pictures and representations audiences are presented with impact how people see themselves and others (e.g., Coronel & Federmeier, 2016; Meeks & Domke, 2016). The question of identity is also a motivator for analyses of workforce diversity—significant because content power exists at the level of journalists, owners, and editors (Cottle, 1998; Picard, 1985).

Studies of content are overwhelmingly unanimous in their findings that women are less present in the news than men (cf. Allern, 2001; Armstrong, 2004; Baitinger, 2015). Women are less prioritized in national and television news than in local news (Armstrong & Gao, 2011), tend to feature more in the tabloid press than in the broadsheets (Dan & Iorgoveanu, 2013), and figure more often in feature and entertainment stories than in hard news journalism (Allern, 2001; Eide, 2001). While female politicians have been observed to seek the media’s attention less than male politicians (Aalberg, 2009), when featured, they are more paraphrased, while men are quoted directly (Aday & Devitt, 2001). Female candidates are also more often scrutinized for their personalities than for their politics (Devitt, 2002) and are more easily “scandalized” than men (Pollack, 2009). Moreover, women are often stereotyped within gender parameters (e.g., De Swert & Hooghe, 2010; Save & Hagen, 2015).

Studies of production suggest the reason for these differences is that the majority of hard news reporters are male (Aday & Devitt, 2001; Armstrong, 2004; Devitt, 2002)—a gender balance also reflected in digital newsrooms (Molandstø & Lamark, 2015)—and that female reporters tend to be assigned to softer news issues (e.g., Desmond & Danilewicz, 2010). Others find that the gender of the reporter is inconsequential to gender imbalances in the news (Craft & Wanta, 2004; Liebler & Smith, 1997; Zoch & Turk, 1998). In fact, a focus on gender equality in the newsroom does not necessarily lead to diverse reporting of political candidates (Beam & Di Cicco, 2010; De Swert & Hooghe, 2010). Hence, whereas gender imbalance has been firmly established, partly explained by the organization of news production, few studies have investigated how structural features impact the ability of journalism to achieve gender balance in news output.

Diversity in the News

Questions concerning how the structural framework conditions of journalism impact the diversity of media systems overall are, however, abundant. To ensure that a maximum number of voices can be heard, media structures need to provide a multitude of units where voices can be circulated (Picard, 1985). Ownership and competition in media markets are, therefore, often investigated as part of media pluralism analyses (e.g., Benson, 2009; Humprecht & Büchel, 2013; Sjøvaag, 2014). The same is true for the way newsrooms are organized, how resources are distributed, and how media work is routinized (Dahlstrøm & Hognestad, 2016; Gans, 1979; Reich & Godler, 2015). Local newspapers’ contribution to overall media diversity has particularly been subject to investigation (e.g., Alexander & Cunningham, 2004; Mathisen & Molandstø, 2015; Nielsen, 2015), not least because of the lack of substitutes for local news (Bridges, Litman, & Bridges, 2002; Gentzkow & Shapiro, 2008). For this reason, state subsidies to newspapers are often tied to circulation, frequency of publication, and market position (Ots, 2009). Not only does local content tend to increase with falling circulation figures (Sjøvaag, 2014), circulation is also associated with content quality (Lacy & Fico, 1991) and advertising revenue (Allern, 1996; Furhoff, 1967; Rosse, 1980), reflecting differences in subscribers and audience sizes (Van Cuilenburg & van der Wurff, 2007). While research as well as policy assumes that ownership has an impact on content (Bagdikian, 2004; I ytringsfrihetens tjeneste, 2000-2001), studies of ownership structures nevertheless diverge on the extent to which concentration impacts external media diversity (e.g., Baker, 2007). And while the impact of press support can primarily be seen in the financial performance of newspapers (Picard, 2003), as a policy measure, it aims to encourage distribution, availability, and journalistic quality (Ots, 2009).

Because a purely market-driven media tend to underproduce certain types of content that are critical to deliberation (Horwitz, 2005), regulation aims to provide the necessary structural conditions to enable the kind of channel variety that can ensure diversity in the media landscape overall (Baker, 2007, p. 15). Public service broadcasting charters, therefore, often contain provisions stipulating ethnic and gender diversity in newsrooms (Horsti & Hultén, 2011; Ross & Padovani, 2017). While niche media that target marginal audience segments such as ethnic, political, or socioeconomic minorities perform diversifying functions in media systems, specialist groups can also assume elite, powerful positions (Gabszewicz, Laussel, & Sonnac, 2002) that still leave audience segments unaccounted for, such as children and youth segments, the elderly, women, and local audiences. Media regulation seeks to mitigate the effect that advertising-driven media can have on economically marginal consumer groups (e.g., Lund, 2007; Søndergaard, 1999). Media diversity ideals assume that a fair distribution of voices is important to counter such structural imbalances in social, economic, and political power in society (Jeffres, Cutietta, Sekera, & Lee, 2000). However, structure has rarely been considered when gender balance in the news is considered. We, therefore, seek to find out which, if any, structural properties of the Norwegian media system contribute to above-average performance of female representation in news content.

Manual Content Analysis

A manual content analysis (Dataset 1) was performed to ascertain the distribution of male and female voices in the news on a representative sample of 75 news outlets in Norway (see Online Appendix 1), covering 33% of the total editorial output of the national media landscape. The stratified sample was selected to include a proportionately relevant number of news outlets reflecting the distribution of publications along the structural variables in the news landscape overall, including ownership and publication frequency, circulation and publication level, geographic distribution, market position, and press support (cf. Krippendorff, 2004; Lacy, Robinson, & Riffe, 1995). The manually coded dataset includes six national newspapers and the online news editions of two national broadcasters; six metropolitan newspapers; and 63 local papers published in all regions of the country. The sample is predominantly made up of local newspapers, reflecting the strong presence of local publications in the Norwegian newspaper structure, 48 of which (64%) receive direct press support. A total of 21 newspapers (34%) are independently owned. The rest are affiliated with larger newspaper corporations. While the sample is heterogeneous—the analysis comparing newspapers with varying formal characteristics—the nature of those characteristics underscores the structural variables operationalized in the analysis. As we are looking for structural explanations for differences in gender representations, comparing disparate newspaper characteristics is a necessity, accounted for by the scale of our sample.

The manual content analysis was conducted on a sample of 5,544 news stories collected from the week of October 5-12, 2015. A consecutive week in October was chosen as it does not contain predictable seasonal anomalies (cf. Blach-Ørsten & Willig, 2016). All news items collected during the week in question were coded. While our data collection was digital, the choice was made to code print editions of local newspapers, to account for the generally low number of online stories published by low-frequency newspapers on a weekly basis. Because the coding unit was names appearing in the news, numbers differ from newspaper to newspaper. This necessarily disproportionate distribution is accounted for by including a large set of outlets. Essentially, as our interest was in gender distribution along structural variables, both national, metropolitan, and local newspapers had to be included in the study. In this sense, all coding units (gender occurrences) are equally informative (Krippendorff, 2004, p. 113).

The coding was performed by seven different trained coders (media students at the Master’s and Bachelor levels). The analysis was based on a codebook consisting of 12 mutually exclusive content variables (politics, economy, social issues, crime, culture, sports, lifestyle, accidents, infrastructure, science and technology, weather, and “other”; Krippendorff’s alpha = .80); 14 mutually exclusive source variables (politicians, public officials, organizational representatives, experts, public relations [PR]/spokespersons, journalists, sports sector representatives, business sector representatives, cultural sector representatives, legal system representatives, clients/victims, children/family/friends, ordinary citizens, and what was termed “unsung heroes”; Krippendorff’s alpha = .71); a mutually exclusive location variable noting the geographical focus of the story (local, regional, national, and foreign; Krippendorff’s alpha = .85); and a male/female gender variable (Krippendorff’s alpha = .95). Alpha reliability scores were acceptable for the content, gender, and location variables. Tentative conclusions are acceptable for the source variable (Krippendorff, 2004), as alpha is a conservative measure (Neuendorf, 2002). Reliability coding was sampled from one of the metropolitan newspapers in the sample (Bergens Tidende), from October 5-6, and performed on 100 items for content, gender, and location, and 131 items for sources.

Coded items include news reports, press clippings, and op-eds, as well as all content in the births/deaths/marriages section of the newspapers. Voices or sources in the news were operationalized in terms of visibility or presence (cf. Raddum & Veum, 2006). Designating voices to one of the 14 source categories involved asking “who speaks here?”, and “who are they speaking as?” Because newspapers were defined as gatekeepers to an arena of public discourse, and because local newspapers have more community-oriented profiles than national elite media (Engan, 2016; Hanusch, 2015; Hatcher & Haavik, 2014; Richards, 2013), a wide definition of “presence” had to be employed. As well as coding mayors, athletes, and spokespersons interviewed in news reports, we also coded children greeted in birthday messages, signed event notices, and the subjects of obituaries. This is rooted in the realization that local papers also serve information functions, or as “notice boards” for events in local communities (cf. Hess, 2016). This was largely due to the comparative interest in finding out the diversity of voices in the local news landscape.

Computational Gender Recognition

Dataset 2 consists of 125 newspapers published across the country, including eight national titles, 25 metropolitan titles, and 92 local titles, 69 of which receive direct press support (55%). In all, 11 newspapers in Dataset 2 (9%) are independently owned. Again, we have a stratified sample, consisting of newspapers that had an online edition that could be scraped at the time of sampling. This means we have a higher proportion of national and metropolitan titles in the sample than we do local titles. The imbalance this causes is made up by the large number of local titles included. The 125 news outlets were scraped hourly during the last 3 months of 2015 (October to December) with custom software. A total of 355,533 Hypertext Markup Language (HTML) resources were collected. From these, 320,228 HTML resources were successfully cleaned to yield a “body,” that is, a body text. These represent predominantly news articles or other journalistic content—where primarily advertisements, short messages/greetings, and media presentations (images and videos) constitute the difference. The body section of the resources that yielded a body is hereafter referred to as an article.

A named entity chunking algorithm (Johansen, 2015) was applied to the articles. The discovered named entities were then compared against a list of 2,684 male and 2,581 female names to determine if the entity named was the given name of a person, as well as that person’s gender. In total, N = 551,102 names were found in these articles, of which 158,464 (28.8%) were female names, and 392,638 (71.2%) were male names. As each name was associated with a gender, the samples are viewed as being, per newspaper, Bernoulli trials. Hence, the observed frequency is also the maximal likelihood (ML) estimation of the frequency with which female names occur among all names. That is, the set of samples for any given newspaper is assumed to be a collection of independent, identically distributed (iid) variables. All measurements and the breadth of their corresponding 95% confidence intervals are reported in the online supplement.

A group of outlets is modeled as a normal distribution over the measurement of its members. Hence, a group is characterized by the mean of the model (which corresponds to the average of the maximally likely probabilities of its members) and its standard deviation. The average is not weighted, meaning a small outlet (few samples) contributes to this measure to the same degree as a larger outlet (more samples). The outlets were partitioned according to variables based on the structural indicators chosen for testing (discussed in section “Results”). Many of these groups were diverse, as can be seen from the standard deviation. However, some categories were sufficiently robust to be compared with each other with some confidence, as will be discussed in the next section.

For grouping variables, t tests were performed, as displayed in Table 1. We found significant difference between local and nonlocal newspapers; between newspapers with less than 5,000 and more than 5,000 weekly circulation; between newspapers that were alone in their markets and newspapers in competitive markets; between newspapers with a distribution of 1 to 3 copies per week and between 4 to 7 copies per week; and between press-supported and nonpress-supported newspapers. Of the structural variables tested, only comparing between independently owned newspapers and newspapers with any other form of ownership showed no significant difference.

OPEN IN VIEWER

Table 1.

Summary of t Tests.

Variable	Category tested	Count (C)	Avg.(C)	Variance (C)	Count(-C)	Avg.(-C)	Variance (-C)	Diff. (av.)	t	p
Form of ownership	Indep./nonindep.	13	0.34	0.00	107	0.32	0.00	−0.03	1.69	.09
Distribution level	Local/nonlocal	85	0.34	0.00	36	0.28	0.00	−0.06	6.17	<.001***
Distribution frequency	1-3 eds./4-7 eds.	59	0.34	0.00	56	0.30	0.00	−0.04	3.87	<.001***
Circulation	>5,000/<5,000	51	0.34	0.00	69	0.30	0.00	−0.04	3.95	<.001***
Market position	Alone/competition	78	0.33	0.00	42	0.30	0.00	−0.02	2.21	.03*
Press support	Press support/nonsupport	64	0.33	0.00	56	0.31	0.00	−0.03	2.70	.008**

Note. Comparisons between structural variables. (C) = tested category. (-C) = complementary category.

*

p < .05. **p < .01. ***p < .001.

Gender Distribution in News Sources

RQ1 asks, In what capacity do female voices feature in the Norwegian news media, and on what topics do they speak? Capacity is here operationalized as source function, or the role assumed by the speaker in the text. Topics are here operationalized along content categories present in the news.

Overall, the manual coding shows that 32.9% of sources in the total set of sources (N = 9,131) are female. The distribution of women sources in the news according to the capacity in which they speak (Table 2), shows that the only source function where female voices are overrepresented compared with male voices (51.8% women to 48.2% men) is when females appear in the news as children, or as friends or family of other sources. In all other capacities, women are less represented than men, least of all as representatives of the sports sector (with 13.8% women), or when journalists interview other journalists (where 18.3% are women). Women begin to equal the presence of men when they represent the “unsung heroes” (persons doing a service to the community, usually as volunteers) at 48.4%; “clients,” or “victims” (at 44.6%), or “ordinary” citizens (43.4%). In cases where journalists interview people outside the established power structures, then, women frequent the newspaper columns almost as often as men do.

OPEN IN VIEWER

Table 2.

Gender Distribution According to Source Type.

Source category	Female (%)	Male (%)
Politicians (n = 1,264)	29.0	71.0
Organization representatives (n = 1,504)	27.0	73.0
Children, family, friends (n = 1,062)	51.8	48.2
Ordinary citizens (n = 1,003)	43.4	56.6
Public officials (n = 961)	33.9	66.1
Sports sector (n = 816)	13.8	86.2
Business sector (n = 598)	25.8	74.2
Experts (n = 456)	32.9	67.1
Cultural sector (n = 444)	37.6	62.4
Legal system (n = 373)	21.4	78.6
PR, spokespersons (n = 168)	41.1	58.9
“Unsung heroes” (n = 161)	48.4	51.6
Clients, victims (n = 168)	44.6	55.4
Journalists (n = 153)	18.3	81.7

Note. N = 9,131. PR = public relations.

A chi-square test for independence indicates high confidence in measures of observations regarding the distribution of genders according to the capacities in which they speak, χ²(1, N = 9,131) = 484.965, p < .0001.

When the elites get to speak in the news media, however, males dominate the coverage. In Norwegian politics, women make up 39% of elected county representatives, 28.3% of mayors, and 39.6% of parliament representatives (2015 figures; Statistics Norway, 2017). However, as this analysis shows, women politicians speak in only 29% of cases. Despite the fact that democratically elected bodies and corporate boards are gender regulated in Norway, requiring at least 40% representation of either gender (Proposition No. 97 [Ot.prp. nr. 97], 2002-2003), gender balance in the workforce is not reflected in journalistic coverage. Whereas 67.7% of women contribute to the workforce, women dominate the public sector (consisting of 70% female workers), while men dominate the private sector (representing 63% of workers). Hence, when representatives of the business sector are interviewed, only 25.8% of sources are women. Women occupy 80.5% of jobs in the health sector and 67% of the education sector, something that could be seen as reflected in the 44.2% of sources in social issues stories that constitute female voices (see Table 3). These workforce characteristics may, therefore, contribute to explain differences in gender representations in the news, not least the fact that 65.1% of leadership positions in the Norwegian labor market are occupied by men (all figures 2016). Male sources are, as such, heavily overrepresented in the Norwegian news landscape, particularly when elite, official, and powerful sources speak in the press.

OPEN IN VIEWER

Table 3.

Gender Distribution According to Topic.

Topic	%	Female (%)	Male (%)
Politics (n = 1,557)	17.1	28.8	71.2
Culture (n = 1,375)	15.1	41.3	58.7
Sports (n = 1,276)	14.0	12.4	87.6
Social issues (n = 1,274)	13.9	44.2	55.8
Lifestyle (n = 1,249)	13.7	48.3	51.7
Economy (n = 742)	8.1	27.6	72.4
Infrastructure (n = 555)	6.1	24.9	75.1
Crime (n = 472)	5.2	27.3	72.7
Accidents (n = 212)	2.3	19.8	80.2
“Unsung heroes” (n = 76)	0.8	48.7	51.3
Science, technology (n = 74)	0.8	28.4	71.6
Weather (n = 55)	0.6	41.8	58.2
Other (n = 214)	2.3	32.2	67.8

Note. N = 9,131. “Percentage” refers to the distribution of topical categories in the sample.

Gender Distribution, Topics

In terms of how men and women are distributed across topics in the news landscape overall (Table 3), women figure more as sources or voices in the soft news categories (cf. Reinemann, Stanyer, Scherr, & Legnante, 2012), such as culture (where 41.3% of the sources are women) and lifestyle (48.3%), than in the hard news categories. Where we are least likely to see women featuring as sources in the news is in sports stories (only 12.4% of sports voices are female), in accident reports (19.8%), and in stories about infrastructure issues (roadworks, park developments, construction projects, etc.; 24.9%), science and technology stories (28.4%), crime (27.3%), and the economy (27.6%).

A chi-square test for independence indicates high confidence in measures of observations regarding the distribution of genders according to the topics on which they speak, χ²(1, N = 9,131) = 566.717, p < .0001.

The manual content analysis demonstrates that women voices in the news are never numerically equal to that of male voices (except in the categories of children, family, and friends). Women are, however, better represented in soft news genres than in hard news genres. Women, while more or less fully integrated into the Norwegian workforce, still do not occupy sufficiently powerful positions to reflect a balanced source material for journalists. In many ways, then, the media can be seen to mirror society’s power structures, reflecting gender imbalances in the hierarchies in society.

Given that diversity concerns encourage representativeness in content presented to audiences, the analysis now turns to the question of whether the structural regulations intended to ensure media diversity in the Norwegian system give any indication that framework conditions contribute to diversify gendered sources in the news.

Gender Distribution, Structures

This section outlines the findings from the computational analysis of gender distribution in a corpus of 551,102 names collected from 125 news outlets, analyzing to what extent the structural features facilitated by media regulations impact gender representation in the news (RQ2). Structural conditions are here operationalized as ownership, distribution level, and frequency, circulation, press support, and market position. Overall, the share of women voices make up 28.8% of the computationally coded corpus. The average percent over all outlets in the corpus is 32.3% (SD 5.3%), matching well the findings from the manual analysis (which found 32.9% female voices in Dataset 1).

Table 4 shows the average female representation in newspapers according to their ownership affiliation. In Norwegian media legislation, ownership regulation was, until 2016, regulated to curtail cross-media ownership, ¹ sector and regional concentration, posing a one-third threshold on sector ownership, and a 60% threshold on regional ownership (where the country was divided into 10 regional sectors). As such, the law encouraged ownership diversity on the premise that a diversified ownership structure facilitates diversity in content. Hence, in looking at gender distribution according to ownership affiliation, a distinction is made between ownership form and between corporations.

OPEN IN VIEWER

Table 4.

Female Representation According to Structural Variables.

Form of ownership	Female, M (SD)
State (n = 1)	30.0 (0.0)
Independent (n = 19)	32.6 (4.5)
Foundation (Amedia) (n = 56)	33.6 (4.1)
Corporation (n = 86)	32.5 (5.3)
Schibsted (n = 10)	29.6 (7.0)
Polaris (n = 20)	31.0 (6.4)
Nordsjø (n = 7)	32.9 (4.3)
Agderposten (n = 3)	29.0 (4.7)
Gudbrandsdølen (n = 3)	30.3 (6.5)
Distribution level
National (n = 10)	28.6 (1.8)
Metropolitan (n = 25)	27.8 (4.9)
Local (n = 90)	33.9 (4.7)
Distribution frequency, per week
1 edition (n = 15)	33.1 (4.6)
2-3 editions (n = 43)	34.5 (5.0)
4-5 editions (n = 12)	35.3 (4.4)
6-7 editions (n = 45)	29.7 (4.6)
Circulation
<2,999 (n = 13)	35.2 (2.9)
3,000-4,999 (n = 35)	34.7 (5.3)
5,000-6,999 (n = 18)	31.7 (3.7)
7,000-9,999 (n = 21)	32.2 (4.9)
10,000-19,999 (n = 19)	31.5 (5.8)
20,000-29,999 (n = 12)	29.6 (4.3)
>30,000 (n = 10)	25.9 (2.7)
Market position
Number one (n = 19)	30.6 (5.4)
Number two (n = 20)	31.5 (6.3)
Alone (n = 88)	33.0 (4.8)
Funding model
Press support (n = 60)	33.6 (5.0)
No press support (n = 66)	31.2 (5.2)

While a simple percentage measure shows that public service broadcasting performs below average, with 30.0% female voices, there is no significant difference between form of ownership for the presence of female voices in the news. This is likely due to the heterogeneity of corporate constellations, and the size of the categories used to differentiate titles. For instance, Schibsted, the largest media corporation in Norway, owns the major metropolitan newspapers in the sample—Bergens Tidende, Aftenposten, Stavanger Aftenblad, and Fædrelandsvennen. These share some characteristics in that they have a low percentage of female voices (22.9%, 25.8%, 22.2%, and 22.6%, respectively). Local newspapers that are, in turn, owned by Bergens Tidende (and, subsequently, part of the Schibsted corporation), however, perform better, for instance, Bygdanytt, with 46.3% female voices. Even within corporations, then, newspapers are sufficiently different to suggest that ownership does not have a significant effect on the extent to which newspapers attain balance in gender representation. No corporation and no ownership form analyzed in this corpus perform better or worse than subsequent categories regarding the presence of women in their newspapers’ coverage.

Norwegian media regulation is also geared toward upholding newspaper publication in all parts of the country. The analysis, therefore, investigates the extent to which there is a difference in gender representation that can be attributed to the differences between local, metropolitan, and national distribution newspapers. Table 4 shows that local newspapers do a significantly better job of including women in their news coverage than national titles, and that they perform somewhat better than metropolitan papers. Local newspapers likely do better because local newspapers carry more stories about local community activities (see Table 5 for an illustration of the local content focus of local newspapers in the manually coded sample), and, hence, interview more “ordinary” people and children—categories where women are more represented than in other types of roles (see Table 2).

OPEN IN VIEWER

Table 5.

Topic Distribution in 22 Local Newspapers According to Geographical Focus.

Topic	Local (%)	Regional (%)	National (%)	Foreign (%)
Culture (n = 926)	71.7	11.0	10.4	6.9
Lifestyle (n = 836)	76.2	11.4	10.1	2.3
Politics (n = 805)	49.2	18.8	24.5	7.5
Sports (n = 688)	61.2	23.0	7.3	8.5
Social issues (n = 651)	55.8	18.9	18.4	4.9
Economy (n = 436)	54.6	24.1	18.1	3.2
Infrastructure (n = 406)	81.3	16.7	2.0	0.0
Crime (n = 342)	64.6	26.9	5.6	2.9
Accidents (n = 167)	48.5	37.1	9.0	5.4
Unsung heroes (n = 71)	81.8	7.0	7.0	4.2
Weather (n = 55)	32.7	38.2	16.4	12.7
Other (n = 54)	50.0	14.8	22.2	13.0
Science, technology (n = 50)	26.0	10.0	26.0	38.0

Note. N = 5,487. “Percentage” refers to the distribution of topical categories in the sample.

Although the number of editions per week often characterizes whether newspapers are local, metropolitan, or national in distribution, frequency of publication (Table 4) does not give the same results. Low-frequency newspapers with one edition per week do not do a better job of including female voices in the news compared with other frequencies. This is likely because four of the 15 low-frequency newspapers measured here are national specialist publications with large circulations, which makes low-frequency papers a highly heterogeneous group. While publication frequency is not found to have any significant impact on gender distribution in the news, newspapers with six or seven editions per week do perform slightly worse than subsequent groups, supporting the general impression that larger newspapers tend to have fewer women voices in the news than smaller newspapers.

Publication level, frequency of publication, and circulation numbers also tend to follow each other in terms of newspaper characteristics. When the extent to which circulation has an effect on gender representation in the news is measured (Table 4), smaller circulation newspapers perform better in terms of gender balance in the news coverage than larger circulation papers. There is little difference between papers with a circulation less than 3,000 copies, and papers with a circulation of between 3,000 and 5,000 copies. However, the presence of women in the news seems to decrease with circulation increase. A range between 7,000 and 10,000 circulation performs better than >30,000, while medium-sized papers in the 10,000 to 30,000 range are somewhere in between. As the average circulation of Norwegian newspapers is 5,700 copies per week, newspapers with lower than average circulation could be said to fare better in terms of gender representation than above-average circulation newspapers.

While there are few competitive newspaper markets left in Norway, competition is thought to effectuate information efficiency (Gentzkow & Shapiro, 2008), political accountability (Bruns & Himmler, 2011), and source diversity (Baker, 2007). For this reason, Norwegian media regulation provides direct support to publications that are number two in their markets, aimed at sustaining competition among media content available to audiences. Market position (Table 4)—whether a newspaper is alone in the market or in competition within a market (as either number one or number two in the market)—however, displays only small, albeit statistically significant, differences in gender distributions between newspapers in monopolistic and competitive markets. Supporting competitive news markets does not, therefore, seem to matter strongly for the diversity of genders in newspapers’ content. This may, again, be due to the heterogeneity of the groups. There are more number two papers in capital and metropolitan markets than in local markets—markets that are also characterized by large circulation and high-frequency publications—factors that this analysis indicates do not have positive impacts on the presence of female voices in the news.

Finally, regarding the significance of receiving financial state support, direct press support makes no difference to gender diversity in the content. The intent behind the direct press support is to prevent newspaper closures, to support competition in newspaper markets, and to sustain local and specialist publications thought to contribute to a diversified media landscape. While state support might be expected to yield higher diversity scores, press-supported newspapers are again a highly heterogeneous group, including national distribution papers as well as local and number two papers. The analysis, therefore, finds that press support has no predictive power regarding the distribution of genders in the news.

All in all, the analysis finds that women are equally represented only in cases where ordinary citizens are featured in the news, and only when soft news lifestyle issues characterize the content. While women are on average only included in the news in 32.2% of cases, improvements to this number can only be explained by two of the six structural factors examined here—distribution level and circulation number. Local newspapers with less than 5,000 in circulation perform better than subsequent categories, while national and metropolitan newspapers with more than 30,000 in circulation perform worse. Most structural indicators are, however, too heterogeneous to serve as framework explanations for the observed gender imbalance in the Norwegian news landscape.