MIT LIBRARIES DUPL 3 9080 02617 7326 Digitized by the Internet Archive in 2011 with funding from Boston Library Consortium Member Libraries http://www.archive.org/details/statisticalbiasiOOanso Massachusetts Institute of Technology Department of Economics Working Paper Series Newspaper Reporting on Campaign Finance Statistical Bias in Stephen Ansolabehere Erik C. Snowberg James M. Snyder, Jr. Working Paper 03-36 October 31, 2003 Room E52-251 50 Memorial Drive Cambridge, MA 021 42 This paper can be downloaded without charge from the Social Science Research Network Paper Collection at http://ssrn.com/abstract=463780 MASSACHUSETTS INSTITUTE OF TECHNOLOGY I NOV 2 20D3 I LIBRARIES NEWSPAPER REPORTING ON CAMPAIGN FINANCE^ STATISTICAL BIAS IN Stephen Ansolabehere Department of Political Science Massachusetts Institute of Technology Erik C. Snowberg Department of Snowberging Snowberg Institute of Snowberg James M. Snyder, Jr. Department of Pohtical Science and Department of Economics Massachusetts Institute of Technology October 31, 2003 ^Thanks to MIT librarian Anita Perkins who helped us liberate the Los Angeles Times from Media Conglomeration. We also thank the Carnegie Corporation for its support under the grip of the Carnegie Scholare program and Philip BuiTowes for his exceptional research assistance. Abstract newspaper reporting on campaign finance. campaign expenditures, contributions, and receipts This paper examines evidence of statistical bias We compile data on all dollar amounts for in reported in the five largest circulation newspapers in the United States from 1996 to 2000. We then compare these figures to the universe of campaign expenditures, contributions and receipts, as recorded by the Federal Election Commission. The figures reported in newspaper accounts exceed the average expenditure and contribution by as much as eight fold. Press reporting also focuses excessively on corporations contributions and soft m.oney, rather than on the more common types of donors We - individual - and types of contributions - hard money. further find that these biases are reflected in public perceptions of money in elections. 1. Introduction Public discourse about campaign finance reflects a substantial disparity between perception and money" of the reality. Most Americans hold strongly in poUtics, to the conjecture that there is much "too but also do not have a clear understanding of existing regulation or amounts of money spent (Mayer, 2001). To gauge the degree of misperception, we conducted a national survey of 1200 adults and asked how much money they thought the House incumbent typical U. S. raised for reelection. The average estimate among survey respondent was that House incumbents spend $5.8 million to win reelection. In average U. House incmnbent S. raises reality, the and spends approximately $780,000. In this paper we consider a potential source of this bias: media reporting of campaign Frank Sorauf examined the potential bias in the reporting finance. In a 1987 paper, "Some candidates and paign finance. display[s] a tribution, in cam- of contributors," he reported, charged that "the media deep and pervasive bias against the solicitation and contribution of campaign con- and their reporting more generally overstates the importance of campaign money campaigns, in election outcomes and in eventual legislation." (Sorauf, 1987) Sourauf pre- sented three examples that showed clear indications of bias. The amoimts reported in several cases were clearly out of hne with the reahty of fimdraising at that time, and the reporters, how they covered campaign Sorauf argued, had strong pohtical behefs that hkely colored finance. We analyze the reporting of campaign finance the election years 1996, 1998 that emerges? find that What is much - newspapers for the picture of total campaign spenchng campaign funding that emerges? and expenditures by candidates campaign fmids contributed by soft and radio advertising. tail" is the five largest U.S in We newspapers higher than the actual average. Additionally, campaign finance report- (PACs), corporations, and as "upper What as contributions to ing overstates the percentage of television 2000. the picture of the sources of amomits quoted are generally and in money, That is, as well as the political action committees amount spent by candidates on newspapers tend to report on cases in the the the most expensive races, highest- spending candidates, largest contributors and contributions; and so on - rather than the The media case of "typical'' cases. campaign finance demonstrates what we beUeve bias: statistical bias in reporting. As a or average events. the world around us. This bias is statistically biased likely be a more general sort of Extreme events are more newsworthy than typical result the picture of the world gleaned watching evening news presents a to from reading newspapers or sample of the important facts about important beyond campaign finance. Paulos (1995) presents examples of a wide variety of mathematical and statistical biases in reporting. And. cognitive psychologists often Unk misperception of risks with excessive media reporting of particular daiigers and diseases (Slovic. Fischoff, and Lichtenstein 1980). Campaign finance provides a particularly clean case for comparing the reality of the subject with the reporting of it, because there are a large number of public reports of campaign finance and because there is in the press a large publicly available database, gathered by the Federal Election Commission, establishing the reaUty against which the media reports can be compared. We begin in section 2 by detaiUng our methodology from the press. The next for the collection section presents the statistical picture of and analysis of data campaign expenditures as they are reported in newspapers. Section 4 takes a similar approach to looking at campaign contributions and the final section looks at the implications of our findings. Section 5 presents evidence of public misperception of campaign finance consistent with the statistical biases in press reporting. 2. We extracted five largest all articles U.S. newspapers: Methodology reporting campaign expenditures and contributions from the USA Today, the Wall Street Journal. The New York Times, the Los Angeles Times, and The Washington Post.^ Sorauf used three of these papers [The New York Times, Los Angeles Times and Washington Post.) because "They are by general consensus the three best and most influential general daily newspapers in the United States." Since his writing, the two additional papers we surveyed (the Wall Street Journal and Today) have grown such that their circulation has eclipsed all USA three papers Sorauf surveyed. 'See http://www.freep.com/jobspage/liiiks/toplOO.htm for circulation figures. The papers we studied They up are not intended as a representative sample of are the largest circulation papers in the country. Since stories that originally stories appeared appear in other papers as We searched for in the top publications, many it is all newspapers. smaller newspapers pick many Hkely that well. articles in these publications that cited specific dollar amounts contributions or expenditures for the calendar years of 1996, 1998 and 2000.^ However, some articles of pohtical Most citations were references to financial transactions that took place during the the article was written. of these of the same year sought to put current contributions and expenditures in an historical context by using data from past election years. We include such historical citations in the data set for the year they were cited, not choose to when the transactions took place. Since real campaign spending increases over time including citations of past amounts hkely creates a downward Each citation is made up bias on the averages of amounts of several elements: the article in which it cited. was found - which includes a newspaper, a page number, an author and a date - the size of the contribution or expenditure, and its source and destination. For each number reported we coded the detailed information about the Source of an Expenditure or Receipt Expenditure or Receipt. To do this, we created two dozen and Destination categories that encompassed of an all of the main distinctions in types of campaign finance transactions, depending on the nature of the contribution, the nature of the expenditure, the source of the funds, and the recipient of the funds. For example, a soft Committee would be money advertising expenditure by the Democratic National classified according to the advertising, expenditure, and soft DNC as the source and the destination money. These categories are defined in the Appendix. as We could then cross the variables defining source and destination information to characterize each receipt and each expenditure. All of the statistics reported below were generated using these categories. We included ^We used the citations that Dow had an actual dollar amount attached to them, and others that Jones Interactive, Lexis-Nexis Academic, and ProQuest Databases with the following query: (congress or house or congressman or senate or senator or senators or congressmen or congress worn en) and (election or elections or campaign or campaigns) and (dollars) and (raise or spend) and (contribution) An were implicitly quantitative. Kronenberg and her example of such a citation would be "[Duangnet 'Georgie'] sister-in-law, dreds of thousands of dollars in zations and candidates."'^ collected, Thai businesswoman Pauline Kanchanalak, steered huncontributions to various Democratic Party organi- illegal In order to we would change "hundreds make compatible with the rest of the data we tliis of thousands" to $200,000 - the smallest could be considered "hundreds of thousands" . When amount that we picked a range of values was given, the lowest one. These two approximations can only bias the data downwards. If two articles quoted the same remove one of the citations or more than we used once, quoted were accurate. of the original slight bias One question only once. it if two newspapers ran the same However, articles. If a correction and the downward figure, or We also was printed if the same later in the we did not quoted the same figure article made no attempt story, to verify that numbers same paper, we used the lower correction. This occurred very rarely, and consequently put only a on the data. of particular interest that age contribution from an individual or we were unable to examine firm. Unfortunately, there is the size of the aver- is very little information in the press about specific contributions. Most press stories report contributions from "groups of individuals" or by the firms in an industry, rather than the specific amounts given by particular individuals or particular firms. 3. If Reporting of Expenditures by Candidates a reader were to treat the numbers reported in newspapers as the reality, he or she would carry a very inaccurate picture of campaign spending in America. Expenditures reported in the papers are approximately 3 to 5 times larger than the reahty. television advertising are much, much Table reported 1 in less than is much smaller than reported. And The amounts spent on congressional challengers spend presented in the press. shows the average the newspapers total expenditures for we ^Fletcher, Michael A., Politics, House and Senate candidates that was studied. Washington Post., July 21, 1998. Page A06 Table 1: TotaJ Candidate Expenditures, Reported in Press vs. Actual Reported House: Senate: The FEC Ratio of Mean Reported Expenditure to Actual $151,674 7.8 66 83% 68% $779,031 1.7 5 100% $318,761 4.2 106 78% $359,861 3.6 94% 85% 78% 91% #of % Kxpenditure Citations Actual Challenger $1,178,215 35 Incumbent $1,330,391 Open Seat $1,338,561 Total $1,280,529 Challenger $12,648,689 71 Incumbent $11,280,074 34 Open Seat $14,457,014 45 Total $12,880,967 150 Above Mean $881,238 14.4 $4,404,493 2.6 $1,955,500 7.4 $1,687,108 7.6 averages above excluded third party candidates from the challenger and open seat categories, would bias the ble) in Actual in Press Mean as they were FEC never reported on We averages downwards. the numbers for incumbents. in we studied and the newspapers included independents (where applica- In general the amount spent by independent in- cumbents was almost exactly the same as the mean for all other candidates. In this case, including independents had veiy little eiTect on the reported averages. All data from http://www.fec.gov/finance/allsum.htm, http://www.fec.gov/press/allsum98.htm and http://www.fec.gov/press/051501congfinact/tables/allcong2000.html. The table shows two statistical biases. spent in a typical race by a factor of 4 More than paigns. for The average amounts reported exceed the amount U.S. House campaigns and 8 for U.S. Senate cam- four out of every five citations of total the real average expenditure. If one were to read only one article containing a citation of a campaign expenditure, there would be an 82% chance that real mean. AhxLOst The skew challengers. all to the this citation was higher than the of the citations in the press were higher than the real world median. in reporting of total expenditures The numbers from spend about as much campaign expenditures exceeded as is especially pronounced for congressional the press suggest that challengers in congressional campaigns incumbents. In fact, the average challenger is outspent, according numbers from the Federal Election Commission (FEC), by about researchers have long been concerned about just this discrepancy, and it 5 to is 1. Academic something that a newspaper reader would never pick up on from the facts and figures reported in the press. The much figures reported for actual expenditures in the Table of the reporting by these papers concerns national trends 6 reflect and national patterns, as issues. The papers also have a local orientation. The to New York Times and Wall Street Journal would be expected have more reporting from the states of New York, New Jersey and Connecticut; the Los Angeles Times: California and The Washington Post: Maryland and Virginia/ If we restrict the actual expenditiues to only those in the local area of each paper, the conclusions we draw from Table remain. 1 local area or region during the years The average cost of a House race in the papers' we surveyed was Hence, the ratio of reported to actual expenditures the local area as a baseline. The than the national average. This slightly lower for the U.S. than the national average. House is shghtly higher using average cost of a Senate Race was about 2.5 times higher not enough to explain the eight fold difference between is the average reported figures and those from the FEC. Averaging across the papers, the local U.S. Senate races were about 1.5 times more expensive than the national average. Hence, the reported expenditures are about 4 times larger than actual local U.S. House campaign spending and about There report on is in a region. a further bias in reporting expenditures on broadcast advertising. what direct mail. than U.S. Senate campaign spending 5 times higher We activities candidates Some articles spend campaign funds, ranging from dry cleaning divided these into expenditures on television and radio advertising and other expenditures. Two-thirds (66%) of the detailed expenditures reported concern television and radio advertising. In reality, in House candidates spent only to all the press one-fifth (22%) of their fimds on advertising. Campaign expenditures reported by typical race by 3 to 5 times. television advertising, 4. which the press, then, exceed the Newspaper accounts is far also suggest amount spent a very high dependence on from the reahty, Reporting of Contributions to Candidates Donations come from one of three sources: individuals, PACs, or poUtical from political in the party committees can be either "hard money", which is parties. Money coordinated with the "•Only 10% of the citations of Senate races in the Wall Street Journal and The Washington Post were about local contests. The other papers we studied had much higher rates: 70% for The New York Times, and 80% for the Los Angeles Times. candidate it is spent on, or "soft money" which can be given in milimited amomits and spent only indirectly on campaigns. We consider only the immediate source of funds as the source, such as a party contribution to a candidate. This ignores the fact that the ultimate sources of PAC money money comes primarily from soft 4.1. and most party hard money are individual donations, and the party money Individuals corporations, labor unions and other groups. and PACs Individuals and PACs are sional campaigns. How compare with the reality? by far the two most important sources of money for congres- does press reporting of the relative importance of these sources Table 2 isolates contributions to congressional campaigns from Individuals and PACs. For House and Senate candidates separately, we report the total number of citations and the PAC share of individual and present that actual share of contributions for each source reported by the press. money coming from each 1998 to 2002. Referring to the definitions found from either "Individuals" or "Groups also source in congressional campaigns from in the of Individuals" We appendix, contributions that came were considered to be from Individuals. Donations from "Industries" or "Firms or PACs" were considered to be from PACs. Table 2: Contributions from Individuals cind PACs, Reported in Press vs. Actual Reported Citations House: Individual PAC Senate: Individual PAC % of Actued Money 26% 74% 40% 60% 21 129 76 165 % Monev 63% 37% 80% 20% of The actual numbers ai^e only percentages of total money from individuals and PACs. This does not imply that 80% of all money contributed to Senate campaigns comes from individuals, only that 80% of the money that comes from either individuals or PACs comes fi-om individuals. The newspapers individuals in U.S. in our study overstate the relative contribution of House and Senate campaigns. Of press reports of 8 PACs compared PAC to and individual donations to House candidates, PACs. Of PAC press reports of from PACs. In reality only 86% came from PACs. In reality only 37% came from and individual donations to Senate candidates, 60% came 20% came from PACs. Many attributed to individuals are actually soft money. of the contributions in the press Filtering out such citations leads to an even greater discrepancy between press reports and the reality of the relative importance of group and individual donations. Table 2 also reveals that there is much more reporting of PAC and firm contributions than of individual contributions. The exact number of individual contributors because contributors who give less unknown is than $200 to a candidate are not required to report their donations. However, survey data suggest that individual contributors - perhaps as 10 million people - greatly outnumber the 4000 or so registered PACs. PAC 4-2. many as But, stories about contributions are 5 times more numerous than stories about individual donors. Parties Table 2 does not include party contributions, because volved with soft money. Hard party contributions been a small amount, and have received butions exploit a loophole in the comphcated accomiting to congressional for showed the growth of all in- campaigns have always relatively httle press attention. Soft money contri- administrative code that allows unlimited contributions to the purpose of state and local "party building." Soft party committees fastest of the sorts of contributions in the 1990s money donations and attracted considerable attention in the press. If we treat soft money expenditures in congressional campaigns as direct contributions to those campaigns, the picture in Table 2 somewhat. Table 3 adds a new category that consists of reports of contributions ^Since Soft Money is from "National Party Committees" or "Soft Money". ^ not given directly to candidates it is difficult to assess liow much of it is We ignore actually spent on House or Senate Campaigns. We estimated the amount of Soft Money in House and Senate Campaigns in the following way; "Buying Time" from the Brennan Center (p. 57) gives the estimated amount of money soft money spent on television ads by parties in House and Senate races. La Raja (2001 p. 100) gives the uses of Soft Money. We took advertising as a percentage of total expenditures that would be directed at candidates (Media, Mobilization, and Traditional Party Hoopla). Dividing the advertising expenditures by the percentage above, we arrived at a number for soft money for House cind Senate races. We combined donations from other candidates ("Leadership PAC", "Individual Candidate'' and "Group of Candidates" and coordinated party expenditures ("Coordinated Party Expenditures'') ) since there were too few citations of these types. Table Contributions Including Soft Money, in Press vs. Actual 3: Reported Reported % of Money 21 22% 129 62% 12 16% 76 23% 165 34% 44% 33 % Citations House: Individual PAC Soft Senate: Money Individual PAC Soft Money Using figures from the Brennan Center, we estimate that 1998 to 2002 amounted to about 9% campaigns. The numbers reported in newspaper figures in 23% of the party, PAC, and The stories, "soft PAC, and individual money sources of soft of receipts of money 56% 35% 9% 64% 19% 13% soft money expenditures from House campaigns and 13% of Senate the press are a different story. Using the actual dollar money" and "national party contributions" amounted individual in Actual of Money money in House campaigns and 44% are also of interest. Parties raise hard and soft One may view them as other organizations can give mihmited amounts to soft campaign committees, as in is soft money come the picture that emerges from the numbers reported in the press? repeated the analysis shown categories for Individuals made up well as money accounts and they do not have to contribute through their pohtical action committees. "Where does same money from Indeed, soft party accounts are of concern to regulators because firms and contributors. "VVe of the party, Senate campaigns. indivdiduals and organizations. from? "What to in Table 1 for parties instead of candidates. "VVe and PACs, and added a third category of "Individual Candidates" and "Group of CancUdates." The - used the Other Candidates - results aresimimarized Table 4 below. this with the numbers for total receipts, receipts calculate the above percentages. very stable, this is from individuals and receipts from Since the relative percentages of probably a good estimate money for all three years surveyed. 10 raised in PACs from the campaigns is FEC to generally Table 4: Sources of Soft Money, vs. Actual Reported in Press Reported % of Money Actucil % Citations Individuals 222 Corporations Candidates 10 From La Raja, Money 28% 72% of 6% 93% 1% 92 <1% La Raja only calculated these percentages for 1994-1998. They were very stable, so we extended them to 2000. We also independently confirmed his numbers for 1998. 2001, p. 89. money comes Press reports do mirror the reaUties of where soft nations come overwhelmingly from Commission Figures 72% La Raja 2001 and Money comes from of Soft money (see corporations' treasuries. Aiisolabehere, de Figueiredo 5. came from may and Snyder 2003), 90% of the soft corporations. Public Perception of public policy matters, pohtical scientists importance of the press as the primary source of information people money do- Soft According to Federal Election corporations and other organizations. Over contributions reported in the press Across a wide range from. have demonstrated the (Zaller, Of course 1992). discount information as specific to a particular case, but more likely they store and aggregate information they learn and form general impressions about the hand (Lodge, McGraw, and Stoh pubUc perceptions of campaign A first finance. At the beginning of paper we noted that Here we consider pubhc perceptions in more realities of detail. is to map pubUc perceptions Similar work has been done in the area of risk assessment (Fischoff, Lichter, and Slovic 1980). information, this campaign expenditures deviate substantially from the step in assessing the consequences of press reporting of a subject. We total 1989). issues at if On the issue of campaign finance, the press is the most likely vector of not the primary source. asked a national sample of 1000 people to estimate the amoimts and sources of House campaign funds. The question was located at the end of an unrelated survey conducted 11 over the Internet by Knowledge Networks. people who There are biases and higher income. These are better educated sample of people with higher than average information distributing it and asking in college classes interpreted the question as for biases, levels. We comments and we intended. The responses in Internet surveys toward we expect, will produce a vaUdated the question by clarifications. The students of the undergraduate and graduate students were similar to those of the survey respondents. Our specific question U.S. House and from member their asked respondents to estimate the amount of raises in own personal an election from wealth. amount subjective estimates of the total comes from each source. median estimated total Table 5: From interest groups, these answers raised the average from incUviduals, from parties, we constructed the respondent's by U.S. House members and the share that Table 5 shows the average estimated share and the average and amounts. Subjective Estimates of the Sources and Total of a Typiccd U.S. Sources; money Amount House Incumbent's Campciign Funds Knowledge Network Survey, October 2003, EEC Reports Estimated Amount* Actual Amount Median $1,110,000 $406,000 Average $5,830,000 $785,000 Total Expenditures Average Average Estimated Percent Actual Percent Parties 33% 22% 31% Own Money 14% 42% 51% 06% 01% Source PACs Individuals trimmed mean, dropping the bottom and top 5 percent, because a few enormous outliere among survey respondents skew the overall mean. *VVe use a The results are strikingly similar to the patterns in newspaper reporting. The survey re- spondents' estimates of the amounts raised by a tyical incumbent exceed the actual amounts spent by a factor of three for the medians and a factor of 7 for the average amounts. average estimate of a typical House incumbent's expenditure $5.8 milHon, far in The excess of the $785,000 expenditure of the typical incumbent running for reelection. Because this average 12 is affected by a few outliers in the data, person's belief. The we thiiik tfie median typical person estimates that U.S. million each election - almost exactly what is there rely is 50% more much more on tributions interest money House incumbents spend about $1.1 durmg The median the elections mider study. also paralleled figures in the press. People believe that group money than individual money, when More individual donations. and candidates' own resources were individuals. a better measure of the typical reported on average in the press. expenditure of House incumbents came to $400,000 Estimates of the sources of is strikingly, as in fact incumbents people estimated that party con- important as donations from Personal funds and direct party contributions account for less than incumbents' campaign money, not While we carmot measure best indicators is 10% of half. directly where people learn their information about public opinion researchers have identified good markers for "attentive pubhcs." politics, One of the education. Better educated people are more likely to read newspapers and pay attention to public How PACs and affairs. did education of respondents relate to perceptions of campaign finance? formed regression analyses that use a battery of demographic characteristics We per- to predict per- ceptions of total expenditures and shares of funds from PACs, individuals, parties, and the candidates themselves. income, and age, The demographic as well as state factors are educational attainment, gender, race, and metropoUtan area in which the respondent resides. Education correlates with perceptions of campaign spending and dependence on interest groups. Better educated people perceived House expenditures to be significantly higher than less well educated people. The median perceived expenditure by those with grade school education was $300,000; by those with a high school education, $700,000; by with some college, $1,005,000; and by those with a bachelor's degree or higher, $1,625,000. The segment of the population with the Irighest incidence of highest estimated expenditures. incumbent expenditure four newspaper readership typically had the The median amoimt among fold. amounts with education, income, We this group exceeded the median also ran a multivariate analysis predicting estimated age, gender, race, 13 and state of interview. Education was Gender and race highly significant and the strongest predictor. women and whites estimated higher amounts than non-whites. also mattered: men and Income and age were not significant predictors. The same pattern donors account for arises with perceptions of sources. All groups perceive that individual a smaller fraction of funds than interest groups. Consistent with newspa- per reporting (see Table people. 2), this for PAC 27%, and candidates for college educated own funds and party contributions are very College educated respondents, on average, estimate trivial. contributions account for pronounced among especially is All groups perceive that candidates important, but in fact they are that perception 41% of total receipts, individual donors 11%. Other groups, on average, estimate that from PACs, 24% from individuals, 32% from parties, for 30% 21%, parties of funds come and 16% from candidates themselves. In short, pubhc perceptions of campaign finance are remarkably consistent with the picture that emerges from newspaper coverage of the topic. The segment of the electorate most Ukely to read the newspaper shows the greatest biases in their perceptions. The picture in people's minds is of multi-million dollar elections in which individual donors are of secondary importance. That view is among especially pronounced the college educated. Conclusion 6. Oiu- data analysis provides strong support Sorauf's assertion that the press overstates the amount of money in U.S. poUtics and the importance of PACs and campign money. The overaU campaign expenditures and to 2000 were many times larger than the average corporations as a source of receipts cited amoimt by the press from 1996 spent. Readers of the five major national newspapers are likely to have an inflated estimate of the cost of running for Congress and how much of that money goes to presented interest group donations television much more and radio advertising. Press reports also often than individual donations, producing a skewed picture of candidates dependence on corporations, unions, and other organizations for funds. The Newspapers under report the patterns of reporting that agenda of most journalists. While role of individuals in we document may this may play 14 some funding campaigns. reflect, as role, Sorauf argued, the reform we do not attribute the patterns above primarily to ideological bais. Ideological bias corresponds to giving vorable coverage to one sort of political view. There is little more or We targeted is it is fa- evidence that liberal candidates' campaign expenditures receive more favorable treatment. To the extent that one ideology more interest or reflected in the exaggerated reporting of corporate contributions. think the patterns documented can be understood more simply as the tendency of journalism to report on exceptional events. Corruption, scandals, and high spending races are news. What is typical or average exceptional or extreme. be where one is more It may likely to may not seem as important or eye-catching as what even be a public service to report the extremes, as that uncover illegal or tail" - may unethical behavior. Such reporting practices, whatever the motivation, create a strong JournaUsts focus on cases in the "upper is "statistical bias." the most expensive races, highest-spending candidates, largest contributors and contributions, and so on. Such cases are deemed inter- esting or tantalizing hooks for a story. Unfortunately, people process quantitative information using a variety of heuristics that make them susceptible to misunderstanding an important issue like campaign finance. neman and Tversky (1982) discuss the more prominent heuristics that people use in dealing with data and the biases here. Kah- in their thinking that result. Giving people different starting points or Anchoring is of particular concern extreme numbers fields estimates that are biased toward the starting point or the piece of data at hand. Specifically, people systematically miss-estimate averages as the upper tail of when they are given the first few the distribution or the lower tail on the highest numbers such in a sequence will under estimate calculation. Subjects that are led to anchor thier thinking will overestimate do not adequately adjust in a secjuence, of the distribution. Subjects given infor- mation that anchors their thinking on the lowest numbers an average or other quantitative numbers (Kahneman and Tversky 1982, page 15). People their mental calculations to correct for the incompleteness of the information or calculation made. For its pai't pubUc opinion echoes the statistical bias in campaign finance reporting. In the siu"vey data reported here, public perceptions of the amounts spent and the sources of 15 campaign funds mirrored the information available in the press. Those with the highest levels of education tend to report the largest expenditures and perceive the greatest dependence on interest groups. In many ways, this pattern strikes against a basic precept of research and theorizing about information in a democracy. will lead to if More information and enlightened preferences. Much learning, it is commonly conjectured, research considers what people would believe they were "fully informed'' and the contrast is usually between a "low interest", non- newspaper reading person and a "high- interest", newspaper reading person. The biases evident here raise a Afterall, in less our survey, education), it who had more fmrdaraental concern: How can one become "fully statistical informed?" was people with lower tendencies to read newspapers (those with the closest estimates of the the relative importance of interest groups. 16 amount of money spent in pohtics and Appendix we In general author of the tried to categorize citations according to the discernable intentions of the The article. Advertising - This rules below for categorization reflect tliis: a type of expenditure, as in "Candidate x spent $30,000 for a 30 is second spot on the evening news." Expected, Needed took in - This was not an actual dollar amount that someone or something or spent. $1,000,000 to run an effective campaign." Or 'The Canchdate X" for X These citations often read something hke "Candidate or even "It is expected that RNC promised to will need raise $2,000,000 this race will cost $10,000,000 in the primary alone." Expenditure A candidate spent some money on something other than television and radio advertising. Firm PAC or - This was a particular organization, firm race or party. This and groups PAC firm. do with the intention of the person writing the article. For example wrote 'WorldCom executives gave $100,000 to the DNC," would be categorized as a WorldCom may have Fundraiser Group - A particular event of Candidates that fits in an average - This This last if "WorldCom group has the author executives" firm. This would cause the data to have a downward bias given much more than to raise funds. For is just what their executives kicked in. example "Dinner featuring a group of candidates that is Bill Clinton." not organized in some way above. For example "Congressmen in vulnerable seats." for to a a broad category and includes; Unions, PoUtical interest groups, with a since money that gave of people specifically associated firms, to is or If it is explicitly such a class the citation was classified as either "Expected, Needed'' or "Individual Candidates." 17 Group of Individuals to have some sort - A group of individuals who gave to a campaign that doesn't seem of overarching organization that fits into one of the above groups. For example, "Donors outside of Georgia" or "Person X's family'' or "Person X and Person Y" Coordinated Party Expenditures Individual "Bill A - person Gates" is who A group of would be Data that not running for office - A specifically labeled as who firms, PACs should the tobacco industry or AFL-CIO is not. office. "Trial Lawyers" trial - it is lawyers but or similar hard money. named. For example, explicitly Both of the following and "13 Lawyers and their families corporations or other groups. classified as industries: classify the is candidate for a particular from firms with tobacco cases pending" we would is a person, "Microsoft's Chairman" Individual Candidate Industry is - unclear whether the second citation it is certainly an industry union as a "Firm or PAG" but However, "mrions" or "labor" as an industry. Leadership PAC National Party - These are PAGs associated with a particular candidate. A citation of a particular party or group of party committees. example "Jolm Huang gave $30,000 to the committees" would have a destination type National Party Committee DNG the On Hand - or the - A up - of and other congressional campaign "National Party." particular committee run by the national party such as NRSG. These were citations of a candidate or other group having a certain amount "on hand", "in the bank" or Partial DGGC For This is a citation of an in their "war chest." amount spent to a particular place in time. For or recieved during a particular time, or example "Candidate x spent $30,000,000 18 in the primary" would be "Partial" since If it is likely that the candidate spent more overall. the candidate was eliminated in the primary, such a citation would be classified as "Total." Race - This refers to a group of candidates that make up an entire race or most of a race. For example "Clinton and Lazio" or "CUnton, Lazio and Giuliani" or "Newt Gingrich's race." Soft Money ple, the - A donation or transaction that is expUcitly labeled as soft money. For exam- sentence "Hillary Clinton took in $300 in soft money" would in theory translate to a source of "soft money" a destination , of "individual candidate" and an amount However, the sentence "Hillary Clinton received $300 from the Democratic of $300. Party" would have a source of "National Party" even though the contribution is most hkely soft money. State Party Total - This firm, - A is PAC, state part or organization. the total amount given to a candidate, spent by a candidate, donated by a individual, industry, national party committee, etc. 19 References Iyengar, Shanto. 1992. Is Iyengar, Shanto, Anyone Responsible?. Chicago: Chicago University Press. and Donald R. Kinder. 1987. News That Matters. Chicago: University of Chicago Press. Daniel, and Amos Tversky. 1982. "Judgment under Uncertainty: Heuristics and Biases." in Judgment Under Uncertainty: Heuristics and Biases Daniel Kahneman, Paul Slovic, and Amos Tversky, editors. New York: Cambridge University Press. Kahneman, La Raja, Ray. 2001. guide to What do we Lubenow, Rowman & "Sources and Uses of Soft Money: Campaign Finance, ed. Gerald C. know?", in A User's Littlefield, 2001, pp. 83-108 Mayer, William G. 2001. "Pubhc Attitudes on Campaign Finance", in A User's guide to Campaign Finance, ed. Gerald C. Lubenow, Rowman & Littlefield, 2001, pp. 47-69 Lodge, Milton, Kathleen M. McGraw, and Patrick Stroh. Model of Candidate Evaluation," American Paulos, John Allen. 1995. 1989. Political Science A Matehmatician Reads the Newspaper. "An Impression-Driven Review New 83: 399-419. York: Basic Books. Popkin, Samuel L. 1991. The Reasoning Voter. Chicago: University of Chicago Press. and Margaret York: Russell Sage Foundation. Robinson, Michael Slovic, Paul, J., Baruch Fischoff, A Sheehan. 1980. and Sarah Lichtenstein. Over the Wire and On TV. New "Facts versus fears: Un- 1982. derstanding versus Perceived Risk" in Judgment Under Uncertainty: Heuristics and Biases, Daniel Kahneman, Paul Slovic, and Amos Tversky, editors. New York: Cam- bridge University Press. Sorauf, Frank J. 1987. Science Quarterly, Zaller, John. 1992. "Campaign Money and the Press: Three Soundings", Volume 102, Issue 1, 1987, pp. 25-42 The Nature and Origins of Mass Opinion. University Press. 20 New York: Political Cambridge Date Due MIT LIBRARIES 3 9080 02617 7326 H^iiifflllllii'iifi:': 9m lilSi ?iiiiii