Roman Larson

Major Grant Proposal: February 2013

How Political Advertisements Delegitimize the Government

Overview
Since the invention of television in the early 1950s, televised political advertisements have played an increasingly larger role in American political campaigns. Talk to a citizen from a battleground state such as Wisconsin or Ohio, and she would probably tell you about the seemingly unending political advertisements in the days leading up to a presidential election. Moreover, that voter would probably allude to the fact that political advertising has become more negative in recent timesa trend confirmed by countless think tanks, media outlets, and scholars. One big question that emerges from this trend is: so what? Americans can, after all, simply turn off the television and ignore televised political advertisements completely. This research project will attempt to shed some light on the so what question. In particular, I will focus on two relationships: first, political advertising and perceptions of corruption of government and the political process (i.e. campaigning); and second, political advertising and trust in government and the political process. I hypothesize that as negative political advertising increases, perceptions of corruption will increase, and levels of trust will decrease. The distinction between corruption and trust is important due to the causal relationship between the variables. For example, if the individual believes the government is corrupt, then she also probably believes that the government is not trustworthy. However, an individual may not trust the government, but at the same time believes that the government is not corrupt. Moreover, if the campaigns themselves are responsible for increased perceptions of corruption, candidates who win office may face a trust deficit with their constituents, exactly the opposite intended outcome of a campaign. Negative political advertisingparticularly the kind that directly attacks another candidate may create an environment where governing is more difficult because citizens trust their elected leaders less. I will attempt to untangle the complicated relationships involving trust in government, perceived corruption of government, trust in the political process, and perceived corruption of the political process. Evidenced by numerous court cases and countless pieces of legislation on the national and state levels, Americans have shown an interest in reducing corruption and even the appearance of corruption. If political campaigns are delegitimizing the governmentby increasing the perceptions of corruption amongst citizensthen governing may be more difficult for political leaders. Thus, examining the connection between political advertising, perceptions of corruption, and the trust Americans place in government will have important ramifications for both the political world and the field of political research. The following summarizes the research questions and deliverables of the project: Research Questions: How do political advertisements influence the perceptions of government corruption and levels of trust in government amongst citizens? Do different types of political advertisement, such as a negative issue ad or a negative character attack, alter the size of the effect? Does the effect differ when separating corruption of and levels of trust in the political process (i.e. campaigning) versus the government?

Roman Larson

Major Grant Proposal: February 2013

How does this effect vary in states that receive significantly more political advertising? How significant is the effect in national campaigns, congressional campaigns and gubernatorial campaigns?

Deliverables: An analysis of advertising effects that uses statistical techniques to support the conclusions drawn A unique dataset to the political field which codes television advertisements from 1950 to 2008 at the level of second-by-second

Literature Review and Significance

The research idea is rooted in the history of campaign finance reform. The federal government has tried numerous times over the last six decades to regulate money in politics, in an effort to reduce perceptions of corruption and increase trust in the political process. To some degree, lawmakers have been successful: limiting individual contributions, for example. In other cases, the Supreme Court has undone regulations passed by Congress, as was the case in 2010 with the Citizens United ruling. Ultimately, however, the goal of any campaign finance reform effort has been to reduce corruption or even the appearance of corruption in government. My research project will focus on perceptions of corruption instead of outright corruption because that is the same standard used by the Supreme Court. Since Buckley v. Valeo, the Court has maintained that even the appearance or perception of corruption is enough of a reason to regulate campaigns. If negative advertising does increase perceptions of corruption in government or reduce trust in government or the political process, then the implications for campaign finance will be interesting. The conclusions drawn from this research paper may shed light on the effectiveness of campaign finance reform in relation to advertising effects; perhaps future efforts to reform campaign finance will need to adopt new approaches. My literature review focused in part on campaign finance reform relating to political advertising and perceptions of corruption. Nathaniel Persilys Perceptions of Corruption and Campaign Finance (2004), for example, illustrates the idea that public perception of corruption has little to do with campaign finance reform. He asserts that other factorssuch as presidential job approval ratingdrive perceptions of corruption. Discussing the relationship between campaign spending and advertisements, Sobermans Campaign Spending Limits and Political Advertising (2007) offers an explanation for the increasingly partisan campaigns between Republicans and Democrats. This study in particular relates to the proposed research because I will be examining differences in perceptions of corruption between states that experience the partisan nature of a presidential campaign and those states that do not. Overall, the studies presented in this grouping suggest interesting ideas that helped motivate parts of this project. Political corruption is inherently linked to campaign finance reform, which is also linked to spending on political advertising. The proposed research study will continue to build on those connections by directly examining the connection between corruption and advertising. Political scientists (i.e. Iyengar and Ansolabehere (1999), Hartmann (2012), Lau (2007), and others) have been studying for generations the effects of political advertising on numerous outcomes such as voter turnout, polarization, candidate preference, the duration of a political advertisements effect, and many more. Gerber, Gimpel, and Green, for example, provide a compelling overview of the duration of advertising effects. Their field experiment showed that the effect of an advertisement is extremely short: the impact of an advertisement on a voter after one week is close to zero. Focusing on
2

Roman Larson

Major Grant Proposal: February 2013

advertisings effect on voter turnout, Professor Iyengar, for example, argued that negative advertising has significant demobilization effects. In general, many studies in political science examine a particular effect of advertising. My project will contribute to that field by researching an unexamined potential advertisement effect. Considering the relationship between effects already studied and the effect I desire to study may produce even more compelling implications related to campaign finance reform. Overall, political advertisements play an important role in campaigns, and they have received much attention, particularly more recently with the dramatic increase in negative advertising. My proposed study will build on two areas of political science: campaign finance reform/corruption and advertising effects. The proposed study connects the two groupings in a meaningful way, which will generate new conclusions about the goals achieved, or lack thereof, to combat perceptions of corruption with campaign finance reform.

Design and Methodology

I will create an advertising database using a second-by-second coding scheme to compare to existing data on corruption and trust. Collecting Advertising Data (Elections after 2000) The collection of advertising data for this project will involve two approaches. The first approach will focus on elections after 2000, using data from the Campaign Media Analysis Group. The second approach will focus on elections before 2000, using data from the Julian P. Kanter Political Commercial Archive. Additionally, quality checks will be incorporated into the coding process where appropriate to ensure the datasets reliability. In 1997, Kantar Media formed the Campaign Media Analysis Group (CMAG) to collect and monitor political advertising. The group has produced data used by academics, national media outlets, and politicians for all elections since 2000. CMAG data provides comprehensive and reputable political advertising data by monitoring nearly all media markets in the United States. CMAGs technology allows the organization to know the instant a political advertisement has aired, and its technology then tracks the advertisement until it is no longer on the air, providing valuable information like how often the advertisement was aired and how much money was spent. CMAG compiles its data in numerous ways. The perhaps most useful feature for political researchers is the visual storyboard produced for each advertisement, which includes a transcription of the advertisements audio. The University of Wisconsin maintains datasets for the 2000, 2002, 2004, and 2008 elections, while data for elections after 2008 is maintained by Wesleyan University. Obtaining the dataset from the University of Wisconsin will be simple, as The Wisconsin Advertising Project only requires the submission of a few forms and a small payment in exchange for its entire dataset. Data for the 2010 and 2012 election cycles from the Wesleyan Media Project is currently unavailable. Wesleyan cites contractual obligations with CMAG that prevent releasing the data for four years after a particular election. After Wesleyan declined to release any data to me, I contacted Kantar Media to work directly with the organization on obtaining raw data for the 2010 and 2012 elections. They quoted me a price of around $100,000 for each election years data, a figure that makes it impossible to work with data from the 2010 and 2012 elections. The robust CMAG data will therefore only be possible to work with for the 2000, 2002, 2004, and 2008 elections.

Roman Larson

Major Grant Proposal: February 2013

Collecting Advertising Data (Elections before 2000) As previously mentioned, CMAGs robust coverage of political advertisements only includes post-2000 elections. No similar compilation of advertisements exists for elections before 2000. Nevertheless, a less robust analysis of elections before 2000 may still be accomplished with the assistance of the Julian P. Kanter Political Commercial Archive at the University of Oklahoma. While much of the study will focus on elections from the past decade (due to the availability of information from CMAG such as how often a particular advertisement was aired), a portion of the study will examine historical trends. This portion of the project will focus on analyzing a small random pool of advertisements for the 1952 to 2000 elections. This small random pool will help to establish if something of interest or significance is happening that is similar or different to the observations of elections after 2000. The University of Oklahoma makes the advertisements in its archive available for review gratis if done so in-person. Its collection is large, containing more than 90,000 commercials. I have spoken on the phone with the archives curator about logistical details to conduct research in the archive. I plan to spend two to three weeks building a small data set coded with my unique coding scheme described below. Best efforts will be made to ensure that a random sample is drawn, but the dataset coded from the archive will be less robust than the one coded from CMAG data. These limitations are unavoidable because advertisements before the 2000 election were not carefully monitored, and the archive is built with donations from news organizations and individuals working with the media. Nevertheless, the archive has agreed to work with me and will allow me to view any sample of advertisements that I select. Analyzing the advertisements from Oklahomas archive will provide a historical perspective to the project. Although there will be significantly fewer variables to analyze as a part of that dataset, the main variable of interest (amount of negative and nonnegative content in an advertisement) will be coded in the same manner as the data obtained from CMAG and the University of Wisconsin. Coding Scheme The coding scheme is unique to the field as it codes advertisements by timing negative and nonnegative content. Whereas most scholarly work has simply coded a political advertisement as broadly negative or broadly positive, this coding scheme will specify the extent of negativity and non-negativity for individual advertisements and allow for a more nuanced analysis of overall trends in political advertising. The following coding scheme is based on a less rigorous coding scheme designed as a part of a research paper relating to this subject that I wrote in autumn 2012 for the course Campaigns, Voting, Media, and Elections. The ultimate goal of the coding scheme is to assign an overall score to each advertisement. The overall value will be determined by first, timing how much of the advertisements content fits into one of the categories and then second, applying a series of rules based on percentages of the advertisements content falling into one of the categories associated with the overall rankings. The following lists possible overall rankings an advertisement might receive: -2: a negative advertisement that attacks or criticizes the opponent; i.e. a character ad that is negative (an attack on an opponent) -1: a negative advertisement that is negative about an issue or the state of current affairs; i.e. an issue ad that is negative

Roman Larson

Major Grant Proposal: February 2013

0: an advertisement that has a mix of both negative and nonnegative 1: a nonnegative issue advertisement 2: a nonnegative character advertisement (i.e. a biographical advertisement, one that does not discuss a particular issue)

Differentiating on these five levels allows advertisements to be distinguished on two dimensions: (1) character vs. issue advertisements and (2) negative and positive advertisements. These differences are important to the overall question of the research project because the effects associated with negative two ranked advertisements may be significantly different from effects associated with negative one ranked advertisements. Additionally, coding the data in this way will allow for several different analytical approaches. For example, I might consider analyzing a subgroup that includes only issue advertisements. I could easily identify these kinds of advertisements by grouping advertisements coded as positive one and negative one. It is important to note that the zero ranking does not reflect advertisements with neutral (i.e. content not matching the other descriptors) content, but rather advertisements with mixed content. Overall rankings assigned to advertisements will be determined using a simple mathematical formula based on the content of an advertisement. Each advertisement will be timed and the amount of seconds spent on negative character, negative issue, nonnegative issue, and nonnegative character, (corresponding with the 2, 1, 1, 2 rankings identified above) will be recorded. Content that does not fit into one of those categories will be recorded under a separate category for neutral content. Neutral time will be subtracted from each advertisement, so data analysis will focus only on content that matches one of the four categories. I anticipate using this neutrality categorization only for the threesecond byline mandated by the Bipartisan Campaign Reform Act. This timing approach will allow for a more comprehensive and accurate breakdown of an advertisements content. Most other research published on ad effects has used a coding scheme that ranks ads as all negative or all positive. In a May 2012 press release, for example, the Wesleyan Media Project indicated that 70% of the advertisements in the 2012 Presidential Election up to that point had been negative. 1 Their requirement for an advertisement to be considered negative was if the advertisement referenced an opponent. As opposed to just categorizing an advertisement as all negative or all positive, this timed approach will allow for unique perspectives into the content of individual advertisements. The second-by-second approach will also allow for advertisements from an entire election to be analyzed in aggregate, more accurately. The following hypothetical illustrates an example of how the coding scheme might be applied. The thirty-second advertisement begins with a ten-second biographical sketch of a candidate running for President. These ten seconds would be coded 2. For the next 17 seconds, the advertisement is a direct personal attack on an opponent. These 17 seconds would be coded as -2. The spot then closes with three seconds of the candidate approving the message. These three seconds would be added to the neutral category and not assigned any value. Thus, the ad would be analyzed as a 27-second ad.

Fowler, E. (2012, May 02). Presidential ads 70 percent negative in 2012, up from 9 percent in 2008. Retrieved from http://mediaproject.wesleyan.edu/2012/05/02/jump-in-negativity/. 5

Roman Larson

Major Grant Proposal: February 2013

The timed content of each advertisement will then be converted into percentages. Overall rankings will be assigned based on these percentages, using a rule of thirds. The following table reflects how overall rankings will be determined:
If timed content is -2 ranking > 66.67% -1 ranking > 66.67% 1 ranking > 66.67% 2 ranking > 66.67% 2 33.3% < ranking -2 < 66.67% and 33.3% < ranking -1 < 66.67% 33.3% < ranking 1 < 66.67% and 33.3% < ranking 2 < 66.67% 33.3% < ranking -2 < 66.67% and 33.3% < ranking 2 < 66.67% OR 33.3% < ranking -2 < 66.67% and 33.3% < ranking 1 < 66.67% OR 33.3% < ranking -1 < 66.67% and 33.3% < ranking 1 < 66.67% OR 33.3% < ranking -1 < 66.67% and 33.3% < ranking 2 < 66.67% ...Overall ranking will be -2 -1 1 2 A weighted value between -2 and -1: (% of ad -2)*(-2) + (% of ad -1)*(-1) A weighted value between 1 and 2: (% of ad 1)*(1) + (% of ad 2)*(2) Description of Ad Type Mostly negative personal attack Mostly negative issue Mostly nonnegative issue Mostly positive personal Mix of negative personal attack and negative issue Mix of positive issue and positive personal

Very mixed ad: includes negative and positive content of both types

While at first glance these rankings seem complicated, the calculations are simple once the correct Excel formula is written. Consider the hypothetical advertisement from above: ten seconds (10/27=.37) would be coded as 2, and 17 seconds (17/27=.63) as -2; three seconds of neutral time would be eliminated from the calculations, hence the 27 in the denominators. Finally, because the ad has ranking -2 for 63% of the time and ranking 2 for 37% of the time, the overall ranking of the advertisement would be zero. These rankings are simply meant to put a final score on the advertisement, similar to how other scholars have scored advertisements. The key difference (and improvement) with this coding scheme is that negative and nonnegative timing within advertisements is specified and the overall ranking is a function of those specific timings. The scale selected for ranking the advertisements is arbitrary and the numbers themselves hold little meaning. Rather, the significance of this scale emerges when examining an advertisements ranking relative to the other advertisements. For example, an overall advertisement that has a ranking of -1.1 would be overall negative, but more focused on issues as opposed to an advertisement that received an overall score of -1.9 which would be an overall negative advertisement focused on character. Thus, the scale is simply meant to facilitate an analysis of the advertisements in relation to one another. Along with the timing of the advertisements, other ad characteristics will also be recorded. These characteristics include the sponsoring candidate or organization, who paid for the advertisement, and if the advertisement was for a primary election or general election. Some of these characteristics will already have been coded in the University of Wisconsin dataset. Ads to Be Coded Depending on the size of the data received from the University of Wisconsin and CMAG, it may not be feasible to code every single advertisement for a particular election. In that case, a random sample will
6

Roman Larson

Major Grant Proposal: February 2013

be drawn in accordance with best statistical practices, and that subset of advertisements will be coded. Presumably, for a gubernatorial election or a congressional election, the number of television advertisements will be smaller, and all advertisements from the selected campaign will be coded. The decisions about random sampling will be made once specific elections are selected for analysis and a review of the available data for that election can be conducted. Corruption and Trust Data The project focuses on comparing the content of political advertisements with perceptions of corruption in and trust of government and the political process. The data for perceptions of corruption and trust will come from the American National Election Studies (ANES) and Gallup polling, and other polls deemed relevant to the project available from databases such as the one maintained by the Roper Center. No survey organization has ever directly asked a question that includes the word corruption. However, questions relating to corruption and trust in government have taken the form of, for example, Do you think that quite a few of the people running the government are crooked, not very many are, or do you think hardly any of them are crooked? 2 In determining which survey questions to include for the data on corruption, I will use a similar approach as Persily and Lammie (2004). They identified questionsrelating to distrust, crookedness, and special interestsfrom polling organizations that relate to perceptions of corruption and trust in government. Although imperfect, the survey data will provide reliable historical data related to perceptions of corruption, which will allow me to make comparisons between states. This will be particularly useful when considering the effects of political advertising in battleground versus non-battleground states and presidential elections versus gubernatorial versus congressional elections. Data for trust is clearer and more widespread than corruption data. Survey organizations have repeatedly asked questions on trust, relating to individual branches of government and even political candidates before they became president. For example, a 1972 poll by State of the Nation asked respondents, How much trust and confidence do you have in our federal government when it comes to handling domestic problems in general: a great deal, a fair amount, not very much, or none at all? 3 Similarly, a poll by Hopes and Fears asked respondents in 1964 Let's look at some specific problems. How much trust and confidence would you have in the way Lyndon Johnson and his administration would handle each of these problems--a very great deal, considerable, not very much, or none at all? 4 Other questions similar to the two examples relating to the political process (i.e. campaigning) will be included in the study.

Persily, Nathaniel and Lammie, Kelli. Perceptions of Corruption and Campaign Finance: When Public Opinion Determines Constitutional Law (June 2004). U of Penn, Inst for Law & Econ Research Paper 04-22; U of Penn Law School. Page 146. 3 State Of The Nation 1972 (Form C), May, 1972. Retrieved Feb-23-2013 from the iPOLL Databank, The Roper Center for Public Opinion Research, University of Connecticut. http://www.ropercenter.uconn.edu/data_access/ipoll/ipoll.html 4 Hopes And Fears, Oct, 1964. Retrieved Feb-23-2013 from the iPOLL Databank, The Roper Center for Public Opinion Research, University of Connecticut. http://www.ropercenter.uconn.edu/data_access/ipoll/ipoll.html 7

Roman Larson

Major Grant Proposal: February 2013

Analytics The analysis of the relationships in the overall hypothesisas negative political advertising increases, perceptions of corruption will increase, and levels of trust will decreasewill include the following: Comparisons between presidential, congressional, and gubernatorial elections; due to higher exposure to certain campaigns like those at the presidential level, the corruption and trust effects may be different. Furthermore, in congressional and gubernatorial races, constituents have much more contact with the candidates, and the race itself is more personal and closer to their lives, so this characteristic may also shape advertisement effects. Comparisons between battleground and non-battleground states. Presidential elections and political advertising at the national level is usually focused on a few swing states. Due to the higher volume of advertising in those states, advertising effects in those stateseven for congressional and gubernatorial elections occurring in non-presidential election yearsmay be different from non-competitive states. Particular attention will be given to non-presidential competitive campaigns occurring in non-battleground states. Comparisons between the trust and corruption variables. Do the two variables move together? Does advertising affect one and not the other? This portion of the analysis will attempt to clarify further the causal relationship between these two variables as I analyze how political advertising affects each one separately. A discussion about differences between trust in and corruption of government and trust in and corruption of campaigns. This section will be important because citizens may trust less or perceive as corrupt the political process due to political advertising, but these effects may not extend to the government. This analysis will involve looking at levels of trust in the federal government when a particular group of citizens experiences a competitive non-presidential campaign. Whether or not specific types of advertisements are associated with different effects. For example, citizens may respond differently when advertisements are primarily about reducing government corruption as opposed to personal attacks.

Much of the analysis will rely on datasets that can be manipulated using statistical analysis software such as Stata or SPSS. My spring econometrics course will prepare me to perform statistical data analysis. The second-by-second approach will also provide advantages to the analysis. First, the approach allows each individual advertisement to receive a more accurate overall ranking because of the secondby-second coding scheme. These rankings can be used to analyze data and relationships in a similar manner as other scholars have done. The particular advantage to the coding scheme, though, is the aggregate analysis that can take place. For example, the percentage of negativity in an entire election can be accurately determined by simply summing up all of the negative seconds that were coded. These exact percentages have never been calculated. Combined with the amount of times that a particular advertisement was aired and seen by citizens, the aggregate analysis of the advertisements will more accurately reflect the negativity, and non-negativity that citizens are exposed to in particular elections. Thus, second-by-second approach allows advertisements to be analyzed using overall rankings of the advertisements and then the aggregated timing data.

Roman Larson

Major Grant Proposal: February 2013

Timeline
The bulk of this project will be completed over the summer with a trip to the University of Oklahoma planned for the end of summer. The timing of the Oklahoma portion of the study will be at the end of summer, so the project will primarily focus on the data from CMAG, the University of Wisconsin, ANES, and Gallup. The project may also be extended into next school year if new questions emerge that require different data. See the following table for an overview: Date April/May 2013 June/July 2013 July 2013 August 2013 End of August 2013 Early September 2013 Task to be Accomplished Work to obtain necessary data from respective institutions Code datasets and begin preliminary analysis Analyze data, draw conclusions, and write up results Travel to University of Oklahoma to code data for elections previous to 2000 Draw conclusions from Oklahoma data and incorporate into written results Continue writing and analysis, plan for further work related to project

Preparation
This research project originated in a research paper written for Professor Iyengars course: PoliSci 120B, Campaigns, Voting, Media, and Elections. The 12,000-word paper analyzed presidential advertisements, employing a coding scheme similar to the one described above (the dataset I built included over 300 advertisements). The writing and research process for that paper exposed me to the data and literature I will be interacting with throughout this project. Additionally, the course laid some of the theoretical groundwork on which my research will be based. Because of the interest I had in the original research paper, I embarked on a directed reading with Professor Iyengar to explore some of the literature related to the research questions posed with this proposal. That directed reading is ongoing, and he and I will continue to discuss relevant literature throughout the rest of the quarter. Professor Iyengars expertise and perspective has been and will continue to be invaluable to this project. He is one of the foremost experts in the field of political communication, and his own work has been included in my studies to help strengthen my own understanding of the broader field related to this project. Our biweekly meetings have served me well as opportunities to discuss questions that have arisen while reading. His commentary and feedback has helped immensely in my own synthesis of the research I have studied. Going forward, this relationship will be maintained, and I will continue carrying out this project with his input. Other relevant coursework in political science I completed includes PoliSci 224L: The Psychology of Communication about Politics in America. The course reinforced and expanded upon many of the ideas presented in PoliSci 120B. Additionally, I gained experience working with Stata in PoliSci 4. Finally, I have taken two statistics courses. During Spring Quarter, I will take an econometrics course to develop my statistical analysis skills. Many of the techniques I will learn in the econometrics course spring quarter may be useful in analyzing the data from this research project. I have also been communicating with the organizations and individuals that I will need to work with to facilitate the data collection and coding process. I have been working with the curator at the University of Oklahoma archive to establish a data collection plan. Additionally, Professor John Geer of Vanderbilt University has provided his own insights on working with the Oklahoma archive. Professor Geer used the archive for a political advertising project he conducted in 2008, and his feedback has

Roman Larson

Major Grant Proposal: February 2013

helped me address some of the deficiencies of Oklahomas archive. Additionally, I have reached out to individuals at the University of Wisconsin Advertising Project, the Wesleyan Media Project, and Kantar Medias CMAG to discuss obtaining and working with the data of each organization.

Budget
The following is a breakdown of my proposed budget: Amount $100 $800 $3100 $1300 $700 $6000 Purpose Wisconsin Advertising Project data Travel to and around University of Oklahoma Stanford university summer housing and food for 8 weeks Oklahoma University summer housing and food for 2-4 weeks Stipend Total

10

Roman Larson

Major Grant Proposal: February 2013

References
Campaign Finance and Political Advertising: The Effect of Broadcasting on Political Campaign Spending: An Empirical Investigation. Burton A. Abrams and Russell F. Settle. Journal of Political Economy , Vol. 84, No. 5 (Oct., 1976), pp. 1095-1108. http://www.jstor.org/stable/1830444 Broadcasting and the political campaign spending Arms Race. Burton A. Abrams, Russell F. Settle. Journal of Broadcasting. Vol. 21, Iss. 2, 1977. http://www.tandfonline.com/doi/pdf/10.1080/08838157709363825 Persily, Nathaniel and Lammie, Kelli, Perceptions of Corruption and Campaign Finance: When Public Opinion Determines Constitutional Law (June 2004). U of Penn, Inst for Law & Econ Research Paper 0422; U of Penn Law School. http://www.pennumbra.com/issues/pdfs/153-1/Persily_Lammie.pdf Primo and Milyo. 2006. Campaign Finance Laws and Political Efficacy: Evidence from the States, Election Law Journal, V5(1): 23-39. http://web.missouri.edu/~milyoj/files/elj.2006.5.pdf Soberman, D. and L. Sadoulet (2007), Campaign Spending Limits and Political Advertising," Management Science, 53(10), 1521-1532. http://mansci.journal.informs.org/content/53/10/1521.full.pdf+html Yanna Krupnikov (2012): Negative Advertising and Voter Choice: The Role of Ads in Candidate Selection, Political Communication, 29:4, 387-413 http://www.tandfonline.com/doi/pdf/10.1080/10584609.2012.721868 Effects of Advertising in Presidential Campaigns: Ansolabehere, S. D., S. Iyengar, and A. Simon (1999), Replicating Experiments Using Aggregate and Survey Data: The Case of Negative Advertising and Turn Out," The American Political Science Review, 93(4) 901-909. http://www.jstor.org/stable/2586120 Gerber, A., J. G. Gimpel, D. P. Green, and D. R. Shaw (2011), How Large and Long-lasting Are the Persuasive Effects of Televised Campaign Ads? Results from a Randomized Field Experiment," American Political Science Review, 105(1), 135-150. Gordon, Brett R. and Hartmann, Wesley R., Advertising Effects in Presidential Elections (August 21, 2012). Stanford Graduate School of Business Research Paper No. 2080; Columbia Business School Research Paper No. 12/33. Available at SSRN. http://ssrn.com/abstract=1881244 Huber, Gregory A., and Kevin Arceneaux. 2007. Identifying the Persuasive Effects of Presidential Advertising. American Journal of Political Science 51(4): 95777. https://journals.cambridge.org/download.php?file=%2FPSR%2FPSR105_01%2FS000305541000047Xa.pd f&code=b7cd846bdc95576eb1a15cd526574854

Appendix: References

Roman Larson

Major Grant Proposal: February 2013

Iyengar, S. , Jackman, S. , Hahn, K. S. and Lim, J. , 2010-06-22 "Polarization in Less Than 30 Seconds: Continuous Monitoring of Voter Response to Campaign Advertising" Paper presented at the annual meeting of the International Communication Association, Suntec Singapore International Convention & Exhibition Centre, Suntec City, Singapore http://pcl.stanford.edu/research/2008/iyengar-polarization.pdf Lau, R., L. Sigelman, and I. B. Rovner (2007), The Effects of Negative Political Campaigns: A MetaAnalytic Reassessment," Journal of Politics, 69(4), 1176-1209. http://lime.weeg.uiowa.edu/~c030111/campaigning/LauSigelman2007.pdf Measuring Media Exposure and the Effects of Negative Campaign Ads. Paul Freedman and Ken Goldstein. American Journal of Political Science , Vol. 43, No. 4 (Oct., 1999), pp. 1189-1208. Published by: Midwest Political Science Association. http://www.jstor.org/stable/2991823

Appendix: References