. . Date: Thu, 22 Sep 2011 13:20:58 -0400 From: Dwight Hines <dwight.hines@xxxxxxxxx> Reply-To: Net-Gold@xxxxxxxxxxxxxxx To: Net-Gold@xxxxxxxxxxxxxxx Subject: [Net-Gold] Newspaper Quality: Local Government Accuracy: Poisson, Binomial, Normal * * Newspaper Quality: Local Government Accuracy Poisson, Binomial, Normal* . . Dwight Hines Peru, Maine September 22, 2011 . . I receive regular emails from a variety of new media organizations reporting on innovative data gathering, data analyses, and the graphic displays to make information easily and cheaply available. Yet, for all the creativity, my concern is about quality, factual reporting that does not waste my time or money reading false or misleading information. I admire and respect the solid, verifiable work by members of Investigative Reporters and Editors, Computer Assisted Reporters, Empirical Legal Studies, and the extensive, free scholarly articles and reviews from SSRN. . I don?t know if you remember when typographical errors were recurrent nightmares that were so common that mathematicians used them as simple examples for teaching different mathematical distributions. Poisson distributions typically gave the best fit and because the mean equaled the variance, that simplified calculations. So, given that you find fact errors in one newspaper article, can you predict fact errors in the other newspaper articles, like you can predict typographical errors? Are there some simple, practical tools that can be developed to measure and evaluate newspapers on quality? Such tools are simpler than ongoing research is finding in predicting where future articles on topics that interest you will be located . (1), and how quality content interacts with consumer expectations . (2a). What is facinating is that analyses of networks are revealing useful information that is not found when the components of the network are examined and tested individually. There is predictive power in the entire newspaper that is greater than what is found when the parts are examined individually . (2b). Errors of fact not only decrease the strength of the *Gestalt* but would cause increased uncertainty and likely misdirection from reality that we need to share as members of communities. . Can we go further and ask if errors of fact in newspaper articles are predictive of a newspaper?s commercial failure? This is a testable question, with no shortage of data from bankrupt, borderline, and successful newspapers for analyses. Of course, interpretations may be complicated by cutbacks in staff that lead to the increasing errors, but the question that needs to be answered remains, how important is journalism quality for commercial success? . A recent example from the *Lewiston Sun Journal* is a good example of fact errors (3). Reading the article you realize quickly that Eric Conrad, Director of Communications for the Maine Municipal Association (MMA), the primary source of the article?s quotes, does not understand ?nuisance? as a legal term. His statements, thus far, are difficult, if not impossible to substantiate with objective data. For example, many of the statements characterize motives of information requesters, many of the statements make it appear that reaching up to a shelf to retrieve a document in response to a request is highly stressful, that requesters interfere with not just busy, but strapped schedules, that requesters don?t use the information, and that requesters make requests just to abuse government officials and employees. . It?s surprising that there are no quotes about how much money has been spent in small towns for computer systems that have not increased information availability or efficiency of services. It is a paradox that the elected officials and municipal employees who tell us how hard their jobs are have difficulty understanding why a plain citizen would need to make multiple requests for information to help understand how our government functions. Face it, our local governments are complex and mysterious ? most of the world still has local governments that are not as effective as ours and function by instilling cold fear in their citizens. Let?s look at the problem of fact errors more closely and see what we can learn. . It is sad that MMA, a government agency with a computer system, PERSONIFY, with over 8,000 queries in it?s database from different member towns asking for help and advice, can not access the legal, managerial, and marketing riches of the database in PERSONIFY. MMA can not respond to public records requests for PERSONIFY summary data. PERSONIFY data breakdowns and analyses can not be generated and are not used for reports or research, or management or marketing, according to MMA attorney Gove?s response to a Maine Freedom of Access Act request (FOAA). It?s not good enough for MMA, with a computer staff of 18, just to keep count of the total number of requests and to watch jaws drop when they tell how much PERSONIFY costs to purchase and maintain in yearly tax dollars. MMA?s failure to have an adequate information infrastructure increases their risks when they are called to court. . The inability of MMA to objectively support their statements is not safe in that the fair report privilege which usually keeps journalists exempt from libel complaints when quoting official sources, does not apply to MMA (4). Yet, the marketplace, composed of readers like me who want accurate factual information, and reasonably efficient government agencies, is likely not so forgiving, especially when factual errors are many and Mr. Conrad and all of MMA are supported by our taxes. Substantial fact errors by MMA in one article, as we see in the *Sun Journal*, increase doubt as to the accuracy in other articles in probabilities and directions and amounts that can be calculated for both the newspaper and MMA (5). . One way to measure accuracy is to count the statements made by MMA in the *Sun Journal*, separating out the compound statements to make them easy to verify or not. Using the Maine FOAA to request documents and other supporting material from MMA for the statements, we find only three of the 30 statements are partially supported: An error rate of 90%. Now, without using Poisson or binomial, this would indicate that the newspaper has quality issues. Taking other information into account, like the *Sun Journal?s* inability to write about how Peru, Maine, not only does not use competitive bidding for substantial contracts (auditor, assessor), but claims they do, leads to the strong possibility that the newspaper is in trouble ? if not ethically, then commercially. The lack of accurate information is a double whammy because it emboldens bad business practices by local governments, like Peru, Maine, to the point where no attempts are made to obtain competitive bids for heating oil. High factual error rates not only cost the *Sun Journal* discriminating readers but will likely cause a loss of undiscriminating readers because the high tax rates for towns with poor business practices reduce reader?s disposable income, and make the towns unappealing to businesses (Among the states, Maine is 50th in attractiveness to business (6). . As you can see by the attached spreadsheet, the 30 statements by Eric Conrad, MMA, are on the left side, with the dates of emails sent as support for the statements along the top. Only 3 of the 30 statements have partial support, weak partial support. Not cited in the article or the emails is the only scientific study that was conducted with Maine municipal officials ? published a month before the *Sun Journal* article. The scientific survey found no mention of FOAA problems by town officials. The survey did find a major concern of officials was the lack of citizen participation (7). Ironic, too, that the publisher of the scientific survey was the Maine Municipal Association. . Traditionally, newspapers did, and some still do, make government practices public and that helped keep government officials accountable. That tradition, I and many others believe, is still legitimate. In fact, it is a tradition that people will pay money to support. A good example of better quality reporting can be found in the weekly *Rumford Falls-Times*newspaper, owned by the same company that owns the daily *Lewiston Sun Journal*. The *Rumford Falls-Times*, based on a rough Poisson, reports more accurately and is a better investment than the daily *Sun Journal*. The former will less likely be replaced than the latter by machines, no matter how intelligent or comprehensive in coverage be those mechanical monkeys and spiders. (8-10) . We need to quantitatively examine relationships beween newspaper quality and newspaper commercial success, and ask how newspaper accuracy impacts community viability. What could be worse for a community than to receive false feedback on government practices and impacts? (11) We would have no fear from the Department of Justice?s Antitrust Division since Ms. Varney left (12). . These are obvious, answerable questions and the answers will generalize to new media. I?d say try Poisson, (13), and in parallel, go for a modified rarefaction on word species (14). . . *References * . 1) Wu, S., et al. ?Predicting Author Blog Channels with High Value Future Posts for Monitoring.? September, 2011, http://www.ssrn.com/abstract=1927096 . 2a) Halbheer, D., et al. ?Optimal Sampling of Paid Content.? June 2011 http://www.ssrn.com/abstract=1923837 . 2b) Dhar, V., et al. ?The Gestalt in Graphs: Prediction Using Economic Networks.? 2010 http://www.ssrn.com/abstract=1500834 Oestreicher-Singer, G. & Sundarajan, A. ?Visible Hand of Peer Networks in Electronic Markets.? 2010, http://www.ssrn.com/abstract=1268516 . 3) http://www.sunjournal.com/ river-valley/story/1073236 . 4) Saharko, P. ?Reporting False Facts?, Summer 2008 http://www.rcfp.org/news/ mag/30-3/lib-reportin.htm . 5) Bliss, C. & Fisher, R. ?Fitting the Negative Binomial Distribution to Biological Data.? Biometrics, 9, 1953, 176-200. . 6) http://www.forbes.com/2010/10/13/best-states-for- business-business-beltway-best-states.html . 7) http://www.umaine.edu/ sustainabilitysolutions/ sustainability_science/ SSI_projects_yr1/pdfs/ Municipal_Official_Survey_TechReport.pdf . 8) Liu, J. & Birnbaum, L. ?LocalSavvy: Aggregating Local Points of View about News Issues.? , WWW 2008, China http://infolab.northwestern.edu/ media/papers/paper10168.pdf . 9) Bruno, N. ?Will Machines Replace Journalism?, Fall 2011, Nieman Reports http://www.nieman.harvard.edu/ reportsitemprint.aspx?id=102686 . 10) Iacobelli, F. et al. ?Finding New Information via Robust Entity Detection.? 2010 Association for Artificial Intelligence http://infolab.northwestern.edu/media/ papers/tmmWPED_20100615.pdf . 11) Gaur, V., et al., ?Intersubjective Meaning and Collective Action in ?Fragile? Societies. World Bank, 2010. http://www-wds.worldbank.org/external/ default/WDSContentServer/IW3P/IB/ 2011/06/23/000158349_20110623174557/ Rendered/PDF/WPS5707.pdf . 12) Varney, C., ?Dynamic Competition in the Newspaper Industry.? Newspaper Association of America 2011 http://www.justice.gov/atr/public/speeches/268742.htm She is missed. . 13) Dallal, G. Poisson Regression. http://www.jerrydallal.com/UHSP/poisson.htm . 14) Collins, M. & Simberloff, D. ?Rarefaction and Non-random Spatial Dispersion Patterns.? Environ Ecol Stat (2009) 16, 89-103. . . .