Methodology

INTRODUCTION | LITERATURE REVIEW | METHODOLOGY | INITIAL FINDINGS
EXPANDING THE MODEL | CONCLUSIONS | REFERENCES | PDF VERSION

The literature surrounding agenda setting is characterized by a wide array of research strategies, everything from the qualitative case study to the longitudinal study to the classical cross-sectional study such as the one used by McCombs and Shaw in their groundbreaking work. The cross-sectional study is widely used in agenda setting studies. It expands upon a case study approach while still looking at a snapshot in time. It provides a more doable option than the time-consuming and more expensive longitudinal studies by Smith (1987), Jordan (2003) and Mortensen and Serrizlew (2006).

This study, like so many before it, including almost half of the significant studies at the local level, will use the cross-sectional approach as its primary focus and first stage. Although not as ideal statistically as the quasi-experimental design used by researchers such as Soroka (2002), Cook, et. al. (1983), Protess, et. al. (1985) and Leff, et. al. (1986), the cross-sectional methodology, including use of a statistical correlation between data points, accounts for almost half of the agenda-setting studies at the local level (McCombs and Shaw, 1972; Jewel and Cunningham, 1968; Kim, Scheufle, Shanaham, 2002).

Still, the other common approach in agenda-setting studies, the case-study approach, while generally far from quantitative, has its merits and will be used in future research in conjunction with the explanation provided by the secondary independent variables (staff size, market saturation, Web presence and ownership) to provide a more robust look at the papers who have a significant impact on policy outcomes. Many cases studies, include examination of particular, and often sensational, issue-related decision-making processes, characterized by major shifts in the composition of the political agenda such as Martin Linsky’s 1986 study of federal policy making. In short, the studies examined exceptional, “outlier” moments in media and politics such as presidential elections or new policies. While useful in that such studies provide insight into exceptional events, they shed little light on the routine work of government or media (Mortensen and Serrtizlew, 2006), but instead shine light on the most significant weakness of the case-study approach, its almost complete lack of generalizability and are beyond the scope of this initial research.

Peter Mortensen and Søren Serrtizlew (2006) noted that most prior studies have focused on the initial phases of decision-making, the agenda-setting phases such as influencing public opinion, without investigating the media impact on outcomes. It may well be that political discussion and public opinion reflect the media agenda. However, while public opinion is critical to the process of policy change and agenda setting, it is not the outcome. Policy change is. To that end, this study examines both ends of the agenda-setting spectrum, media and policy. The intervening public opinion and the political agenda, which may easily be swayed by public opinion (but possibly not swayed far enough to make lasting changes) are not the focus here.

The Dependent Variable: Policy Change

If it is policy change that is really important, then, it becomes necessary to determine how to look at policy change. Examining one specific policy would be to fall into the trap of looking at some sensational or unique item on a local government’s agenda. Examining only the number of policies changed would prove little other than the willingness of the government to make change regardless of reason.

Aaron Wildavsky (1979) agreed that the budget is a representation of “Who gets what government has to give?”. The budget is the focus of the government’s efforts. The budget is a representation in monetary terms of governmental activity. If politics is, in part, a conflict over who gets what, then the budget records the outcomes of this struggle. On the other hand, David Osborne (1993) said, government today consists of a lot of dedicated people trapped in bad systems—budget systems that provide incentives to waste money, personnel systems and civil service systems that are cumbersome and provide few incentives to be more efficient.

The easiest way to examine the budget would be to subtract the previous year’s budget from the following year’s budget. The dollar amount obtained would represent the change from the previous year. However, it would not account for increases or decreases in the overall budget that might be due to outside factors such as routine adjustments, tax increases or population changes. On the other hand, the percentage change used in this study, as opposed to the actual dollar amount, will make routine adjustment easier to identify and non-routine changes stand out.

While Census Bureau, the International City and County Manager Association (ICMA) and National League of Cities (NLC) may propose a model structure for a local government budget, there is no one, consistent template for a budget determining what goes in one line versus what goes in another and no single source for local government budgets. Choosing four distinct factors in the budget, items that appear in most, if not all, local municipality budgets in one form or another helps to ensure internal validity. Because the number and scope of governmental functions influence both revenues and expenditures, only factors common to all regions have been chosen for this study. For example, many Northeastern cities are responsible not only for general government functions but also for public education. However, in other states, education is a state function. Some local school districts have their own taxing authority and are independent agencies. Some cities are required by their states to assume more social welfare responsibilities than other cities. Cities also vary according to their revenue-generating authority. Some states, notably Kentucky, Michigan, Ohio and Pennsylvania, allow their cities to tax earnings and income. Other cities, notably those in Colorado, Louisiana, New Mexico and Oklahoma, depend heavily on sales tax revenues. Therefore, much of the statistical data presented here must also be understood within the context of cross-state variation in tax authority, functional responsibility and state laws (Hoene and Pagano, 2009).

Most city budgets have at least some common categories. They all have a budget line item for “administrative services.” This line is often small in relation to other lines, including costs of the city manager’s staff and council meetings. Most cities included lines for public safety, parks/recreation, economic development and public works, lines that tended to be much larger, more controversial and which generated media coverage. The pieces that made up these aggregate items can also be informative in terms of what news coverage they generated. In public safety, for example, sometimes coverage of police and fire individually may generate coverage while “public safety” per se does not. So, the areas selected for this study partially mirror Meagan Jordan’s work (2003) and partially reflect the accessibility of modern local budgets, obtained from the governments themselves through online resources, and the line items they contain.

Dependent variables

• Change in public safety budget from previous year (continuous) [public safety change]
• Change in parks/recreation budget from previous year (continuous) [PRT change]
• Change in economic development budget from previous year (continuous) [development change]
• Change in public works budget from previous year (continuous) [public works change]

The Communities

The cities or counties chosen for research need to vary in size, geographic region and other characteristics to ensure some generalizability. As Steve Barkin (1987) said in his study of local television news, the emphasis is not on the distinctive characteristics of the cities chosen but on the generalizability of the local issues and concerns. However, they needed to be cities served primarily by one newspaper to avoid the confounding present in a major metropolitan town served by multiple newspapers, television stations and other media. The town’s budget information must be readily accessible, preferably online,[1] making it easy to access not only for researchers but to media outlets and citizens.

The first step in the selection of cities, town, townships, boroughs and villages (to use the terminology of the U.S. Census Bureau) was that they were included in the list of “local governments” — as opposed to states and federal entities. This included 19,494 municipalities in the 2007 list. Of them, 5,221 (26.8 percent) filed their Web site URL with the Bureau. This pool of 5,221 serves as the initial sample of local governments. The cities with URLs on file with the Bureau of the Census represented all sizes of municipalities (from populations ranging from 37 to 8,214,426) and all geographic regions of each state. There did not seem to be any bias in this sample.

This list will be matched with the communities that have newspapers in the EBSCO Newspaper Source Pro database. All of the communities that remain (appendix b) will comprise the next sample and will be examined for any bias.

The Media

The most conventional agenda-setting studies (and many other studies) make use of content analysis of the media in question (Krippendorff, 2004). For this study, the number of stories devoted to the particular topic during the time in question will be determined from a content analysis. Therefore, all of the newspapers being studied must be searchable using the EBSCO Newspaper Source Plus database[2], a database that includes 906 media outlets. Because the database for newspapers allows for repetitive, repeatable searches of numerous local newspapers through a defined period of time, it will be used for this survey rather than searches of the individual newspaper’s Web sites. The method of determining media coverage will follow Baumgartner and Jones (1993), who say, “When we want to know whether an issue is news, therefore, it is not difficult; we simply count the number of articles published in an index of media attention for a given year.” All U.S., non-national newspapers in the database with coverage back at least three years will be included in the study. Use of this database provides a foundation for the validity of this research; other research will be able to replicate the searches using the database. Because researchers working at different points in time and under different circumstances should get the same results, the research should be more reliable (Krippendorff, 2004).

From the Newspaper Source Plus database of 906 media outlets media outlets outside the United States were removed, reducing the sample to 450 searchable media outlets. Any media outlet that had fewer than three years of data online was then eliminated, leaving 241 outlets. Finally, any duplicates or large, regional papers were removed, leaving 162 media outlets that met all the criteria. The percentage of the newspapers kept in the study closely reflected the percentage of the population in that region. As noted in table 5, the largest region, as determined by the U.S. Bureau of the Census (see appendix a), had the largest percentage of papers in the sample and the smallest region had the smallest percentage of papers in the sample.

Table 5: Comparison of U.S. Newspapers and Population by Region

Region	Omitted	Kept	Kept %	Population %
1	33	33	20.4%	19.0%
2	54	38	23.5%	22.9%
3	136	54	33.3%	35.6%
4	64	37	22.8%	22.5%
Non-U.S.	Â	Â	Â	Â
or non-print	456	Â	Â	Â
Total	743	162	Â	Â

See appendix a for a detailed list of regions as identified by the Bureau of the Census.

The search terms revolving around they key words “public safety,” “parks and recreation,” “economic development” and “public works” mirror Jordan and, to a lesser degree, the analysis by Mortensen and Serritzlew or even Baumgartner and Jones who looked at nuclear power, tobacco and pesticides, not the entire national political process. These also represent items that a newspaper would cover not only when sensationalistic events happen but as part of routine coverage. Each one of the primary independent variables will be examined independently to see if they move together, indicating just general coverage of local issues, or independently, indicating that coverage of that one issue area is having more of an impact on policy outcomes. The cause of that difference will be an area ripe for future research.

---

Threats to Reliability and Validity

To ensure the consistency of the measures in this study, and therefore to ensure reliability, each of the variables has been clearly defined and the method used to obtain the data detailed. By using the same databases and sources for the media data and for the budget of each town, potential random error will be minimized. That there is no single source for local municipality budgets is a potential threat to reliability and could question whether the measurements of a test remain consistent over repeated tests. This will be minimized by using one source for the budget of each town, the town itself, minimizing the chance that future researchers would obtain different data.

As is apparent in all statistics, however, just because a study is reliable, measuring the relationship between policy change and media coverage consistently, does not mean it is actually measuring that relationship. Researchers have identified some nine threats to internal validity: history, maturation, testing, instrumentation, statistical regression, selection of subjects, experimental mortality, selection-maturation interaction, and the John Henry Effect (Yu and Ohlund, 2010; Campbell and Stanley, 1963; Shadish, Cook and Campbell, 2002).

Because this study is not experimental in design, uses third-party data, and does not use pre-test/post-test methodology, it is not likely that there will be a testing bias. Using data from the census, EBSCO database and town budgets, should avoid problems caused by selection of the subjects on the basis of novelty such as is often a problem with agenda-setting case studies. Avoiding an experimental design also avoids problems with selection-maturation interaction and the John Henry effect when the treatment itself changes the outcome.

Five of the potential threats, however, are potential threats to this study: selection bias, instrumentation changes, history, maturation and subject mortality. The first, and perhaps most significant, is selection bias. The towns and selection of the media are both subject to selection bias. The media outlets are chosen because they are a part of the EBSCO database, a selective, not-all-inclusive database. Although not apparent on the surface, those newspapers that choose to use EBSCO may have some commonalities biasing the sample. To avoid too small of a sample size, all of the local papers (not the national or large regional papers) in the database will be used in the research, at least initially. From the available newspapers, then, the towns those newspapers serve will be examined. If that town’s budget is not available, it will be eliminated from the study. Since department-level data was not available on municipal budgets from a single source such as the International City/County Managers Association, Census Bureau, the League of Municipalities or any other group, each budget will have to be examined to obtain comparable data. Towns that have their budgets available electronically may share some common characteristics of openness that might influence the outcome of the study but that are difficult to measure. While collecting budget data, control variables will be used to ensure that the final sample of towns represents a representative sample of municipalities in the U.S.

A second threat to validity is the historical aspect of the study. Even a study over one or two years could be subject to history effects. For example, the municipalities studied could have undergone a change in leadership making them more or less susceptible to media influences. For example, the local media might have undergone a change in ownership, influencing the type and depth of coverage on that papers’ staff. To combat this potential threat, the study will be conducted over a short time period. A more longitudinal methodology provides potential for future research.

Third, it is also possible that changes in technology, instrumentation in the vernacular of the academic study, might influence the study. Even over the few years in which this study has evolved, towns have become more likely to post their budgets online and to make more and more material accessible through their own media outlets rather than relying on the mass media. And mass media outlets have become more likely to post more material online on their own Web sites as well as the EBSCO databases. As more material is available through sources other than the local newspaper, it becomes more difficult to attribute any change to the chief instrument of this study, the local newspaper. However, the same information that might be more available to the general public is available to the mass media. And just because it is available does not mean anyone will go looking for the town budget or drafts of policies. The local newspaper still serves to provide people with information they never knew was there.

Finally, maturation and mortality are similar problems that could pose potential threats to this study. For example, a newspaper may improve its online coverage during the time of the study or may change staff unknowingly causing a change in focus for or against coverage of local issues. In addition, a newspaper could actually cease publication during the study forcing it to be eliminated. Those that continue publishing may be more responsive to their readership and may, therefore, bias the outcome of the study.

Because the study makes use of external data, it is not subject to testing bias but could fall prey to threats to external validity since it is not experimental in design (Campbell and Stanley, 1963). As a cross-sectional study and not a case study provides for some randomization and increases the potential for generalizability. By choosing towns and media outlets following the stated criteria and by controlling for some potential external factors, this study should safely allow for the generalization of the sample studied to local municipalities and the newspapers that serve those communities.

Coverage and policy change

H1: Local coverage and policy change will be positively correlated.

Independent variables

• Change in number of stories regarding public safety (police, fire, EMS, public safety, crime) from previous year (continuous) [public safety story change]
• Change in number of stories regarding parks/recreation (parks, recreation) from previous year (continuous) [PRT story change]
• Change in number of stories regarding economic development (economic development) from previous year (continuous) [development story change]
• Change in number of stories regarding public works (public works, water purification, sewage treatment, trash collection, road maintenance, construction) from previous year (continuous) [public works story change]

Keeping mind all the potential problems with reliability and validity, to identify the number of sources in each publication, the EBSCO database will be searched for the terms in question. The online database allows searches for individual media within a defined time period. So, for each of the content analysis independent variables, the database will be searched using the appropriate terms, ensuring consistency between media for each fiscal year. For example, for the public safety independent variable, the database will be searched using “police,” “fire,” “EMS,” “public safety” or “crime.” The number of stories found with any one of those terms will give the value for that variable to be correlated with the dependent variable, looking for the direction and amount of correlation.

Maxwell McCombs and Donald Shaw (1972) found correlations as high as+0.967 between main campaign issues carried by the media and voter’s independent judgments of what were the important issues. When it came to an abstract, national issue such as the environment, one study even found a perfect correlation (+1.0) between the relative salience of the environment’s attributes in the newspaper and among the public (Maher, 1996). Yet these were not the only studies with high correlations. Indeed, it is such high correlations in repeated studies over decades that have led to the validity of the agenda-setting hypothesis. Marc Benton and P. Jean Frazier (1976) found an especially high correlation (+0.81) between the newspaper agenda and the public agenda on the general topic of the economy. Maria Jose Canel, Juan Pablo Llamas and Federico Rey (1996) found that six major concerns on the public agenda matched local coverage with correlations as high as +0.90 depending on the newspaper. Federico Rey Lennon (1998) found a correlation of +0.60 between the public agenda and the newspaper and of +0.71 between the public agenda and television. Toshio Takeshita (1993), looking at four individual areas and the public agenda published some of the lowest correlations in the literature — +0.39. Wayne Wanta and Salma Ghanem (2007), in a limited meta-analysis of 90 agenda-setting studies, found an overall mean correlation of +0.53 and 95 percent confidence intervals of 0.47 ≤ 0.53 ≤ 0.59. Correlation remains the choice of researchers studying agenda setting.

Although prevalent, correlation as a methodology is not without its problems. As pointed out in the introduction, the correlation (whether high or low) may also be due to other factors. The absence of correlation between media coverage may conceal actual media impact if that impact is offset by prior audience differences and the appearance of correlation may be spurious if it merely reflects parallel audience and/or context differences. The role of the media must be taken in context with other potential agenda setters.

Most of the studies with high correlations, however, looked at the correlation between public opinion of what was important with content analysis of the news media. It follows, based on the model, that correlations would be lower as the media became further removed from the outcome. Since policy changes follow the political agenda which follows the public agenda, they would have the lowest correlations with the media agenda. Regardless of how weak it is, it will be significant, showing that media coverage does have some (positive) impact on policy outcomes. Another of the pitfalls of correlation as a technique is that it does not provide a direction. Correlation is symmetrical. While it is the standard method in use for more than 30 years in agenda-setting studies and is used by the vast majority of the studies of agenda-setting except the ones using case study methodology, it provides only clues for causation not determination.

Despite these statistical limitations, correlation remains the accepted methodology for many agenda-setting studies particular when the direction of causation cannot be part of the study (the base model points that causation may indeed be bi-directional and correlation most appropriately allows for this), resources do not permit an experimental approach or time does not permit a longitudinal study. Statistical methodology examining the primary variable and the secondary independent variables will comprise part of the study with some more qualitative work providing a closer look at cases of interest identified in the qualitative work.

Organizational variables

With this study, the most interesting comparison is that between coverage and policy change. However, coverage, as indicated by the number of stories in that area, is not the only variable of interest. As James Lani (2008) says, “Very rarely – if ever – is it the case that only one variable is responsible for values of another variable.” Other moderating variables help to account for the relationship between the policy outcome, the budget, and the coverage. They influence the strength of a relationship between the two other variables. These moderating variables, including staff size, market saturation, online presence, media ownership, and existing relationships with other local media, will inevitably have some influence on policy outcomes (Baron and Kenny, 1986; Tabachnick, 2001).

Size of staff

H2: The size of a newspaper’s staff will be correlated positively with policy change. The larger the size of the staff, the larger the positive correlation.

Independent variables

• Size of staff in second year (continuous) [staff size]

Media outlets tend to have short attention spans (Downs, 1972; Kingdon, 2003). It is no quantum leap to hypothesize that a newspaper that has the resources to dedicate reporters to beats will have reporters that get to know their sources better and, ultimately, produce more in-depth, well-researched articles. With more staff, they can spend more time on a story. Not only will they get to know when things are going well in an area, they will know when things are amiss and can inform the public. Newspapers that can allocate even more than one person, say a team of reporters, photographers and editors, on a specific issue or topic can produce even better coverage, particularly more community-based coverage (Russial, 1997; Morgan, 1993; and Johnson, 1993). To fill the pages, smaller staffs often have to rely on “wire” copy while larger staffs can produce more locally generated copy about local actors and issues.

David Demers (1996) used staff size as part of his structural complexity index, looking at the number of full-time employees (38.25/mean), number of full-time reporters/editors (184.62/mean) and even the number of beats that employ full-time reporters (5.30/mean). Similarly, Lee Becker (1978, p. 108) looked at the number of full-time employees, part-time employees, and even the gender, salary, staff turnover and education of those employees to show that “The number and kind of editorial staffers hired, their training and experience, and the salary given them all predict to press performance.”

Philip Meyer and Minjeong Kim (2003) asked a simple question: “How many news-editorial staff members does it take to produce a viable newspaper?” To answer the question posed by Meyer and Kim and to test newspaper “folklore” that there should be one news-editorial staff member for each 1,000 circulation, Meyer and Kim obtained staff size information not normally released from the American Society of Newspaper Editors and circulation figures from the Audit Bureau of Circulation, excluding newspapers that were not in either list. They found that the folklore is not far from the truth. The mean news-editorial staff rate for 1995 was 1.04 and increased to 1.18 by 2000. Through their regression analysis, the authors ultimately showed that, although the effect size was small, 5.8 percent, a one-unit increase in staff per thousand results in a circulation increase of 7.427 percent. While they acknowledge that increasing staff size alone cannot stop a circulation decline, they do also show that the top 9 percent of newspapers have the highest staff/circulation ratio. “An enlarged news-ed staff creates benefit as well as cost. The investment analysts who see a newspaper as a platform for delivering eyeballs to advertisers in the cheapest manner possible should think about what attracts those eyeballs to the platform” (Meyer and Kim, 2003).

For this study, three methods will be used to determine the size of the staff in the year’s studied. The first will be the Editor and Publisher International Yearbook that lists staff members. This will be supported with the newspaper’s Web site. However, both are problematic. The Yearbook may not list the entire staff, listing only key personnel instead. Since what one staff lists as “key” personnel may differ from what another lists, this may serve as nothing more than a reliable lists of contacts and way to determine if there has been a change in key personnel during the years in question. The second will be the newspaper’s Web site. However, this too is problematic as it is probably accurate for the current year but will not reflect changes in the staff during past years. To that end, this variable may necessitate contacting a top manager to determine the size of the staff during the years in question. A contact in human resources or in the editor’s office will know how many people worked at the paper during the years in question and will ensure that the numbers are comparable between the years in question at least for that paper.

Market Saturation

H3: The higher the quality of the newspaper as measured through circulation, the higher the correlation between local coverage and policy change.

Independent variables

• Market saturation = circulation/population [market saturation]

Entire books have been written and Web sites devoted to the declining circulation of the printed newspaper. Since 1940, the Newspaper Association of America has tracked the circulation of morning, evening and Sunday newspapers as well as the total number of newspapers in the market. In 2003, there were, according to the Newspaper Association of America 1,456 morning or evening daily newspapers in the United States, about 100 fewer papers than 10 years prior. In the same year, there were 6,704 weekly newspapers, down less than 200 from 1996 (the first year tallied). The combined circulation of the weekly newspapers was more than 50 million in 2003 (Newspaper Association of America, 2004). And since 1964, the Newspaper Association of America has tracked readership (circulation in relation to population). Similarly, the Audit Bureau of Circulations reports the official total circulation of newspapers. These circulation figures have become the official numbers on which newspaper success is measured, and they are declining. They are also the circulation figures on which advertising rates are based. The more readers a newspaper reaches, the more than can charge for advertising. A drop in official circulation means a drop in advertising revenues. For example, the Monroe Enquirer-Journal (Union County, North Carolina) has an average daily circulation of 7,666. All other things being equal, it could not charge as much for advertising as the Greenville Reflector (Pitt County, North Carolina) that has a daily circulation of 20,709 according to the Audit Bureau of Circulation. The Greenville paper reaches 13,043 people per day more than the Monroe paper.

However, circulation alone does not tell the whole picture. Monroe has a population of 31,663 and Greenville has a population of 76,058. Therefore, Monroe’s newspaper reaches 24 percent of the population and the Greenville paper reach 27 percent. When population is taken into account, the papers reach about the same percentage of their community. A paper that reaches a higher percentage of the market is more likely to inform and to motivate citizens, to change public opinion and to influence policy change. While it cannot take into account how many people read an individual copy of a paper or how many people read the paper online, market saturation will be used in this research as a proxy for quality and as a way to assess the likelihood a paper will influence a significant segment of the population.

Beyond just being an indicator of how many papers are being sold or what part of the market the paper is reaching, market saturation, too, is a measure of quality. “It makes sense that the larger the number of people buying a newspaper, the larger the number of people who perceive it to have at least a minimal level of quality” (Lacy and Fico, 1991). In their study, Stephen Lacy and Frederick Fico took circulation and city population figures from the Editor and Publisher International Yearbook (which is what this research uses), stratified the sample by monopoly newspapers, competitive newspapers and papers with a joint operating agreement and then did a content analysis on the papers for one week to create a quality index originally developed by Lee Bogart in 1997[4]. They found there was a significant relationship between news quality and newspaper circulation (r = 0.373). “If a newspaper’s circulation is related to its journalistic quality, as indicated by this and previous studies, it is in the managers’ and owners’ interest to invest in such quality” (Lacy and Fico, 1991).

Gerald Stone and his associates (1981) also acknowledged that while a superior journalistic product is associated with a greater circulation level, it is not the only factor. “Quality can be said to account for at least 3.4 percent of the explained variance in total circulation….” They found that, even with community size is partially controlled, the size of the community had more of an impact on circulation than quality. That is why, for this study, market saturation (circulation/city size) will partially control for this variable.

Web Presence

H4: Local newspapers with a Web presence will show a higher correlation between local coverage and policy change.

Independent variables

• Does paper have a Web site with content (not just business information)? (dichotomy: yes or no) [web presence]

According to the Newspaper Association of America, circulation in 2007 was as low as it was in 1977, having peaked in the early 1990s — before the World Wide Web became commonplace (Newspaper Association of America, 2008). The NAA is not the only group to report the demise of the printed newspaper in favor of online media. Philip Meyer predicts that by 2043, daily newspapers will run out of readers. He cites lack of timeliness as the primary cause of declining readership. “The Internet…(gives) seekers of specialized information an increasingly efficient source. Why would you look up yesterday’s closing price of your favorite stock in the newspaper when you can find the last half-hour’s price on the Internet. (The newspaper)…is no longer the most efficient way to appeal to those interests” (Meyer, 2004).

Major metropolitan newspapers including papers such as the New York Times and Washington Post have taken a leadership role in online media, taking advantage of the instantaneousness of it, the ability to provide readers with up-to-the-second updates as well as the ability to provide moving pictures and in-depth coverage with small incremental costs. Local newspapers, however, have been slow to jump on the bandwagon. It was not until around 2000, nearly a decade after newspapers began producing online, that the average weekly newspaper began publishing online, according to one study (Adams, 2007). Only a few years later, by 2004, according to the Newspaper Association of America, more than 5,000 daily, weekly and newspapers had an online presence.

Because more and more users are getting their news online, examining whether a newspaper has an online presence seems a logical extension of this study. By having the local news online, it would be accessible to more people and therefore more likely to influence their decisions. Further, because online news remains easily accessible for a longer period of time than a newspaper which might be recycled the next day, more people than just subscribers might read the stories pertaining to local government. So, newspapers with an online presence will have more impact on policy changes.

While the growth of the World Wide Web may supplement or replace printed editions, particularly the large city newspaper, it would be negligent to say that there is consensus in any part of the industry about the future of the printed edition. Even Ben Bagdikian (2004), one of the industry’s harshest critics, does not believe the local newspaper will go away because of the unique role papers play in the community. “Newspapers have not yet disappeared, nor are they likely to in the near future. … Newspapers have a unique social function that their media competitors do not. These social functions are likely to extend the life and solvency of the printed newspaper and keep it a substantial presence in the media scene for many years.”

While it is a logical extension of the research, this dichotomous variable presents its own statistical problems. As David Garson (2009) says, “it is common to use dichotomies in interval-level techniques like correlation and regression,” using this dichotomous variable (among the others in use in this study) presents potential statistical problems. In this case, there are varying levels of quality of a Web site for a newspaper. The paper may have no Web site (0). Or it may have a Web site, but that Web site contains only information about the business, no content (also 0). Beyond that, the paper may choose to put only some articles up online, perhaps only objective pieces and no editorials (which can have a profound impact on public discussion) or no letters/commentary. For this study, if the paper puts any content up online it receives a 1, not a 0 regardless of what was put up online or when. In effect, this is a Likert-type scale with only two values. This abbreviated scale, which assumes interval data but with ordinal Likert scale items, does not seem to affect Type I and Type II errors dramatically for many statistical tests (Jaccard and Wan, 1996). With all of the non-continuous variables in this study, it will be important to examine the factor loadings to assess whether common loading reflects substantive correlation (Garson, 2009).

Ownership

H5a: Locally owned newspapers will have a stronger correlation between amount of coverage of local issues and policy change than newspapers owned by national chains.

Independent variables

• Is newspaper part of a chain? (dichotomy: yes or no) [chain]

H5b: Local newspapers owned by the same company as the local television station will show a higher correlation between local coverage and policy change.

Independent variables

• Is newspaper owned by same company as local television station? (dichotomy: yes or no) [ownership]

Only a few cities left in America have competitive newspapers. In fact, 99.9 percent of morning papers are monopolies in their own cities and face competition largely from television stations, not other print media. Five chains own the majority of daily newspapers in America and many smaller papers: Gannett, McClatchy, Advance Publications (Newhouse) and the New York Times Company. Smaller group such as Morris Communications, Community Newspaper Holdings, Inc., Freedom Communications, Lee Enterprise, Belo and Scripps, Cox Newspapers, the MediaNews Group and Hearst own hundreds of mid-sized and small papers. Fewer than 275 of the nation’s 1,500 daily newspapers remain independently owned (Bagdikian, 2004; Free Press Action Fund, 2008).

Bagdikian insists that the downfall of the modern media will be corporate conglomerations. Yet other research insists that chain-owned and monopoly newspapers do not differ much from independent newspapers. Demers (1996) showed that corporate ownership can free reporters to focus on writing and reporting rather than just filling pages. “Corporate newspapers publish more local editorials and letters to the editor, and a larger number and proportion of editorials and letters that are critical of mainstream groups and institutions.” Demers examined papers that were incorporated businesses, publicly owned, owned by a chain/group, and papers where one family/individual does not own 50 percent interest. He hypothesized, and showed, that corporate newspapers publish editorials that are more critical of mainstream groups because the papers are more likely to be located in pluralistic communities that contain more criticism of dominant groups and value systems. Further, they are more insulated form local political pressures. Randy Beam (1998) showed that market-oriented newspapers, those with a focus on customer (both advertisers and readers) wants and needs are not less committed to public affairs content or general journalistic excellence. “A strong commitment to identifying and seeking to meet reader’ wants and needs is not tantamount to pandering to readers.” In contrast, Becker (1978) found that “the large papers, those with large news holes to fill, and those which are part of large media corporations, are better newspapers than those with small circulation, small news holes and which are individually owned.”

While Bagdikian’s argument is compelling, often-cited, and criticized, it is also incomplete. He argues that the money-hungry owners make the media conservative, not liberal yet fails to explain why any media outlet would knowingly bias its coverage. He fails to recognize that there is a distinct separation between corporate ownership and the media personnel who create the content of the mass media. Still, his arguments are also indicative of the prevailing public opinion regarding the mass media and are the basis for the hypotheses in this research.

Also at the foundation of this part of the research is the declining number of media owners and owners that now operate print and broadcast media in a single community. The Tampa Tribune owns WFAL-TV, an NBC affiliate in the same market, indeed in the same building. Until it split into two companies, one for television/new media and one for newspapers, Belo owned both the Dallas Morning-News and WFAA, a television station operating out of an adjacent building in the one of the nation’s largest markets. In any given market, formal relationship or not, their relationship is even more “incestuous” as Michael Rosenblum (2008) said. “Almost every local TV news station starts its day by opening up the local newspaper. That is where TV news gets its stories from. They let the local paper do the hard work, and they simply lift the results.” And with the corporate owners looking for ways to save money, having the broadcast media and print media share stories seems efficient and effective even if it does reduce the check and balance on the gatherers of the news.

Control variables

• Population of city in second year
• Region
• Frequency of publication

The primary independent variables and moderating variables should go a long way in explaining the impact of coverage on policy outcomes if not the actual causation. In all studies, some items are beyond the interest of the researcher but inevitably account for some of the change in the dependent variable. Other studies often use variables such as gender, age, race, education or as control variables, variables held constant so as not to influence the outcome. “Utilizing the control variable will do a great deal to silence the critics of your research that may attribute the differences you found to the existence of some extraneous, unidentified, and unaccounted for variable” (Lani, 2008).

This study uses population in the municipalities studied as a control variable. While population is also accounted for in the independent, moderating variable market saturation (circulation/population), it is also important statistically to control for population changes independently.

The Bureau of the Census divides the United States into four regions: Northeast, Midwest, South and West (see appendix a). Because Census Bureau data will be used for population, the same source will be used for U.S. region to see if there are any differences by region.

The final control will be a dichotomous variable regarding the frequency of publication: daily or non-daily.

The Relationship

To investigate the agenda-setting capacity of the local media, this study attempts to correlate what city governments passed as policy in terms of the budget with the actual content of the local media used during the same time frame. While this is not a long-term, longitudinal study, meaning it will be difficult to prove causation, there is so little work in agenda setting at the local level, even a small step is a step in the right direction. With some 400 studies on agenda setting at the national level, it will take more than a dozen studies to build enough literature to determine how, or if, agenda setting works at the local level. By choosing towns and cities of varying sizes and in varying geographic areas and by choosing a variety of cases chosen the findings of the study will be generalizable even when the sample size limits its generalizability. The first part of the findings in this study will look at the relationship between the budget and the newspaper coverage, generally replicating other agenda-setting studies but using multiple policies, not just one case as many others have done.

Despite some ambiguity, researchers over the years have generally accepted that mass media coverage can, and often does, affect the political agenda. However, from a policy perspective, this conclusion does not tell us very much. We do not know much about the strength of the media influence on political decisions. It is not clear whether institutional, political or economic factors magnify or diminish the effects of media coverage (Mortensen and Serritzlew, 2006).

When McCombs and Shaw completed their original study, they used intercorrelations between issues and voter emphasis (public opinion) and found extremely high intercorrelations, +0.967 between main campaign issues carried by the media and voter’s independent judgments of what were the important issues. A similar methodology will be used here, testing the correlation between coverage and policy change in four areas: public safety, parks and recreation, economic development and public works. This methodology, looking at a snapshot of a community at one point in time may prove correlation is symmetrical and does little to prove causation, a weakness of the original studies that was corrected in later studies at the national level and in a Danish study. In their time-analysis study of Danish municipalities over a 13-year period, Mortensen and Serritzlew (2006) went one step further and assessed changes in spending per citizen and media coverage, getting at causation as much as correlation. They found little significant impact of the media on local government budgets over a 13-year period.

The longitudinal nature of their study allowed Mortensen and Serritzlew to look at causation which routine examination of correlation does not, but being unable to determine causation is not the only problem faced when using correlation, an indication of the strength of the (linear) relationship between the variables. The first potential problem depends on the strength of the model, a model tested in various other research projects. The strength of the model should prevent the coefficients being found significant by chance alone. The strength of this study is based in no small part on the strength of the model as substantiated by findings as indicated in the literature review. A second potential problem is also negated by the strength of the model. If other variables not accounted for in the model also cause the dependent variable, policy change, then any covariance they share with any given independent variable may incorrectly be attributed to that independent. Finally, because correlation assumes a linear relationship between the variables, if the relationship is, instead, non-linear, correlation will understate the relationship (Garson, 2009).

Over time, various researchers have found that media coverage precedes public opinion and policy outcomes. However, as indicated the vast majority of the research shows only a relationship, not the direction of the relationship. This is a weakness of most of the studies, including this one. Over the short term, as the model indicates, media coverage can lead to public opinion change which can result in policy change. However, it’s also likely, especially over an extended time period, that policy change can result in media coverage. When researchers study extended time periods, indeed, it is possible that feedback occurs in both directions. While the literature shows that public officials use the media and other sources to get items for potential policy change on their agenda as the policy changes mature, it is just as likely that policy change results in further media coverage.

After examining the relationship between the coverage and changes in the budget, this study will examine other variables that might have an impact on budget changes: newspaper staff size, market saturation, Web presence, ownership and co-ownership with local broadcast media. While these variables are not the major focus of the study, policy change is, they might provide some insights into the factors that cause that policy change beyond just coverage in the local print media.

Table 6: Hypotheses and Variables Used in this Study

Number	Hypothesis	Type of Variable	Variable
Â	Dependent variable	Continuous	Change in public safety budget from previous year
Â	Dependent variable	Continuous	Change in parks/recreation budget from previous year
Â	Dependent variable	Continuous	Change in economic development budget from previous year
Â	Dependent variable	Continuous	Change in public works budget from previous year
Â	Control	Continuous	Population of city
Â	Control	Discrete	One of four U.S. regions
Â	Control	Dichotomy	Daily/non-daily
H1	Local coverage and policy change will be positively correlated.	Continuous	Change in number of stories regarding public safety from previous yearChange in number of stories regarding parks/recreation from previous yearChange in number of stories regarding economic development from previous yearChange in number of stories regarding public works from previous year
H2	The size of a newspaper’s staff will be correlated positively with policy change. The larger the size of the staff, the larger the positive correlation	Continuous	Size of staff in second year
H3	The higher the quality of the newspaper as measured through market saturation, the higher the correlation between local coverage and policy change	Continuous	Market saturation = circulation/population
Â
H4	Local newspapers with a Web presence will show a higher correlation between local coverage and policy change.	Dichotomy	Web site with content (not just business information) or not
H5a	Locally owned newspapers will have a stronger correlation between amount of coverage of local issues and policy change than newspapers owned by national chains.	Dichotomy	Locally owned or not
H5b	Local newspapers owned by the same company as the local television station will show a higher correlation between local coverage and policy change.	Dichotomy	Owned by same parent company or not