Persuadable perceptions: the effect of media content on beliefs about corruption Lucia Rizzica Marco Tonello Bank of Italy Conference on Corruption, Tax Evasion and Institutions Riga, 11 May 2017 Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 1 / 24
Introduction and motivation Corruption is commonly defined as the misuse of public office for private gain (Svensson 2005). It is by nature secretive and thus not observable. We study the individuals perceptions about the extent of corruption in their country 1 influence individual choices on many critical issues (e.g. voting, investment, work) 2 used to build measures of corruption CPI Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 2 / 24
How accurate are perceptions? Perceptions vs. victimization 2014 EuroBarometer: less than 4% of respondents across EU declared to have been asked to pay a bribe in the 12 months before the interview almost 80% think corruption is very widespread in their country Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 3 / 24
What we do in this paper Research questions: 1 Does exposure to corruption related news (corruption salience) affect perceptions about the degree of corruption in the country? 2 Does the news content matter? What content? Preview of our results: (Potential) exposure to news about corruption increases individuals perceptions about the extent of corruption in their country The content of the news matters: non-factual news influence beliefs more than factual ones persuasion rather than learning explains our results Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 4 / 24
Related literature and contribution 1 Media persuasion: Voting: Della Vigna and Kaplan (2007); Barone et al. (2015) Politics, policy, public spending: Strömberg et al. (2010, 2015) Crime, justice: Dahl and Della Vigna (2009); Ouss and Philippe (2017) Family: Chong and La Ferrara (2009); Bassi and Rasul (2017) Our contribution: first on effects on corruption perceptions first to look at online information micro level evidence (individual characteristics) not just exposure to media but content 2 Measures of corruption: Corruption perceptions vs. hard measures: Fisman and Miguel (2007); Olken (2009) Our contribution: identify a novel channel of bias in perception based indices (net of definition issue) Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 5 / 24
Data: corruption perceptions The Italian Survey of Households Income and Wealth (SHIW) We added to the 2014 wave of the SHIW a set of questions about corruption perceptions We obtained subjective probabilities about the perceived extent of: 1 petty corruption 2 bad social norms conductive to corruption 3 effectiveness of measures to contrast corruption Sample of about 1,800 heads of household Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 6 / 24
Data: corruption perceptions The SHIW vignettes Imagine that a citizen refers to a Public Office for some service. Attach to each event the probability (between 0 and 100) of how likely is the event to occur (low values indicate low probability of realization, high values high probability of realization): the public officer hints that he would accept a sum of money, a favor or a gift in exchange for providing the service (Corruption) the public officer hints that the citizen has to ask for the intervention of a friend or acquaintance who works in the Public Administration in order to expedite the service provision (Social Norms) a corrupt public officer who has been discovered, serves a term of imprisonment (Effectiveness of Justice) Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 7 / 24
Data: corruption perceptions Descriptive statistics Corruption Social Norms Justice Enforcement N Mean Sd Mean Sd Mean Sd Occupational status Employees 532 31.48 26.76 45 29.9 18.23 22.6 Self employed 136 32.16 27.46 42.88 30.12 11.10 21.61 Not employed 1137 36.69 27.94 46.36 31.5 14.42 19.03 Education level Higher education 841 32.49 26.39 44.76 30.75 15.29 20.42 Lower education 964 36.45 28.64 46.45 30.88 15.81 21.01 Contacts with Public Offices Lower than the median 965 33.28 26.35 44.65 29.43 14.61 19.6 Higher than the median 840 35.79 28.91 46.68 32.33 16.64 21.89 Total 1805 34.44 27.58 45.59 30.82 15.55 20.71 Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 8 / 24
Data: corruption salience on the media The Corruption News Dataset (CND) Between Jan-March 2014, we built the CND containing daily information on all news items related to corruption: searched on the front pages of 30 national (N) and local (L) on-line newspapers, btw 2 and 4 PM each day online newspapers recorded the number of news items that contained the words: corruption, bribe, misuse of public office, embezzlement Built a measure capturing, for each respondent residing in a given province p, the potential exposure to corruption-related news on day d: News dp = j N news dj + j L news dl cjp>0 (1) Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 9 / 24
Data: corruption salience on the media Newspapers and geographical variation il Fatto Quotidiano La Stampa la Repubblica Il Centro Italia Oggi Il Messaggero L'Unione Sarda Il Tempo Il Mattino La Gazzetta del Mezzogiorno Messaggero Veneto Il Secolo XIX Il Giorno Il Giornale di Sicilia il Giornale Corriere della Sera Il Gazzettino Libero Il Sole 24 ORE Avvenire La Nazione Gazzetta del Sud Corriere Adriatico il Resto del Carlino Il Gazzettino (Venezia) Il Tirreno Corriere dello Sport-Stadio Tuttosport La Gazzetta dello Sport Corriere dell'umbria 6 4 2 2 50 50 48 44 42 41 41 38 35 29 25 25 24 24 23 22 22 22 19 19 19 18 17 14 98 122 0 50 100 150 Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 10 / 24
Identification strategy A graphical illustration (a) (b) (c) (d) (e) (f) (g) Corruption perceptions and news (residuals) -40-20 0 20 40 1/8/2014 1/22/2014 2/5/2014 2/19/2014 3/5/2014 3/19/2014 Date Corruption news Corruption perceptions Notes: Media coverage and corruption perceptions by date. The vertical lines denote, in chronological order: (a) embezzlement cases involving the Piedmont governing body; (b) the so-called Ruby ter trial in which Silvio Berlusconi was accused of corruption; (c) the issuing of the EU Anti-Corruption Report; (d) a case of bribery in the Italian Space Agency and news on the EU Anti-Corruption Report; (e) the appointment of a new prime minister (Matteo Renzi) with statements about the fight against corruption; (f) the so-called Maugeri trial in which members of the Lombardy governing body were formally accused of corruption and embezzlement; (g) the case of corruption discovered in the Lombardy public health system. Residuals from OLS regressions on day of the week fixed effects. Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 11 / 24
Identification strategy Estimating equation Y iedp = α 0 + βnews dp + α 1 X i + α 2 Int i + ϕ e + ϕ t + ϕ p + ɛ iped (2) Y iedp : perceptions about Corruption, Social Norms, or Justice Effectiveness reported by individual i, residing in province p and interviewed by enumerator e on day d; News dp is the measure of exposure to corruption news items; X i is a set of individual characteristics (educational level, occupational status, frequency of contacts with Public Offices); Int i is a set of interview characteristics (duration and if carried out in the morning); Fixed effects: ϕ t day of the week, ϕ p province, ϕ e enumerator. Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 12 / 24
Identification strategy Test for random scheduling assumption Dep. Variable: Newsd,p Female -0.00 0.15-0.03 (0.33) (0.29) (0.30) Age 0.01-0.00-0.02 (0.02) (0.01) (0.01) High School 0.72* 0.32 0.40 (0.41) (0.38) (0.38) College 0.54 0.58 0.66 (0.53) (0.54) (0.58) Employee 0.03 0.50 0.17 (0.57) (0.45) (0.42) Self employed -0.41 0.05-0.07 (0.86) (0.39) (0.41) Civil Servant -0.86-0.88-0.56 (0.62) (0.58) (0.66) POF: at most 5 times py -0.07-0.59-0.18 (0.69) (0.59) (0.53) POF: at most 10 times py -0.16-0.58-0.05 (0.80) (0.62) (0.63) POF: more than 10 times py -0.35-0.32 0.02 (0.66) (0.69) (0.66) Morning -0.26 (0.29) Duration -0.02 (0.01) Constant 11.92*** 13.74*** 13.30*** (1.81) (2.63) (4.72) Adj.R2 0.00 0.22 0.29 N.Clusters 64 64 64 N.Observations 1805 1805 1805 Individual characteristics Fixed effects: province, dow Interview controls Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 13 / 24
Baseline results Are corruption perceptions influenced by exposure to related media content? Corruption Social Norms Justice Enforcement (1) (2) (3) (4) (5) (6) News 0.26** 0.17** 0.21 0.06-0.07-0.11** (0.10) (0.07) (0.15) (0.11) (0.08) (0.06) Female 2.27 0.36 0.97 (1.42) (1.42) (1.21) Age -0.12** -0.19*** -0.04 (0.06) (0.06) (0.04) High School -3.82** -1.29 0.32 (1.69) (1.86) (1.08) College -3.59-0.11-0.08 (2.21) (2.84) (1.69) Employee -2.80-2.25-1.02 (2.18) (2.12) (1.56) Self employed -1.42-1.70-1.78 (2.33) (2.59) (1.67) Civil Servant -4.84** -5.34* 2.40 (2.37) (2.73) (1.93) POF: at most 5 times py 0.33 3.62 2.77 (2.76) (3.44) (1.73) POF: at most 10 times py 3.14 4.61 1.30 (3.32) (4.15) (1.82) POF: more than 10 times py 3.37 6.54 3.31 (3.02) (4.08) (2.00) Adj.R2 0.00 0.28 0.00 0.30 0.00 0.26 N.Clusters 64 64 64 64 64 64 N.Obs. 1805 1805 1805 1805 1805 1805 Individual and interview controls FE: province, day of week (dow) Other results Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 14 / 24
What drives changes in corruption perceptions? Does content matter? 1 Are news informative about the real level of corruption? Corruption perceptions might respond to changes in corruption events (e.g. convictions) reflected in the news (Learning process) 2 News coverage might increase only the salience of corruption Persuasion process through behavioral reactions, e.g. top-of-the-mind, post-truth, negativity bias... Analysis of the informative content of the news Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 15 / 24
The informative content of corruption news Long-term trends in corruption news items, investigations, convictions Convictions/Investigations 0 10000 20000 30000 40000 50000 TANGENTOPOLI 1982 1992 2002 2012 Convictions Investigations News items 0 1000 2000 3000 4000 News items Notes: Convictions is number of people convicted for corruption crimes by year (Ministry of Justice); Investigations is number of crimes recorded by police forces (Istat); news items is the number of news on corruption appeared on the national newspaper la Repubblica by year. The vertical line indicates the outbreak of the tangentopoli scandal in 1992. Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 16 / 24
The informative content of corruption news Convictions as a proxy for corruption events Persons convicted (residuals) -20 0 20 40 (a) (b) (c) (d) (e) (f) (g) 1/8/2014 1/22/2014 2/5/2014 2/19/2014 3/5/2014 3/19/2014 Date Persons convicted for corruption-related offenses Notes: Convictions by date. The vertical lines denote: (a) embezzlement cases involving the Piedmont governing body; (b) Ruby ter trial in which Silvio Berlusconi was accused of corruption; (c) EU Corruption Report; (d) bribery in the Italian Space Agency and news on the EU Anti-Corruption Report; (e) the appointment of a new prime minister (Matteo Renzi) with statements about the fight against corruption; (f) Maugeri trial in which members of the Lombardy governing body were formally accused of corruption and embezzlement; (g) corruption in the Lombardy public health system. Source: Ministry of Justice. Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 17 / 24
The informative content of corruption news Convictions as a proxy for corruption events Panel A (1) (2) (3) (4) Corruption Convictions -0.29 1.52-0.25 0.57 (0.88) (1.22) (0.87) (1.41) Corruption news 0.17** 0.19** (0.07) (0.09) Corruption news Convictions -0.07 (0.13) Constant 40.83** 748.25*** 38.60** 38.35** (16.09) (279.93) (16.42) (16.40) Adj.R2 0.28 0.28 0.28 0.28 Panel C Justice Enforcement Convictions -1.04-0.56-1.06-0.85 (0.68) (1.15) (0.67) (1.00) Corruption news -0.12** -0.11 (0.06) (0.07) Corruption news Convictions -0.02 (0.07) Constant -18.31** 317.78-16.77* -16.83* (8.40) (193.23) (8.84) (8.81) Adj.R2 0.26 0.26 0.26 0.26 N.Observations 1805 1805 1805 1805 Convictions dc Three days moving average of Convictions dc Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 18 / 24
Comparison between factual and non-factual news Factual or event-related news Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 19 / 24
Comparison between factual and non-factual news Non-factual or non-event-related news Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 20 / 24
Comparison between factual and non-factual news Event study (sort of...) (a) (b) (c) (d) (e) (f) (g) Corruption perceptions and news (residuals) -40-20 0 20 40 1/8/2014 1/22/2014 2/5/2014 2/19/2014 3/5/2014 3/19/2014 Date Corruption news Corruption perceptions Notes: Media coverage and corruption perceptions by date. The vertical lines denote: (a) embezzlement cases involving the Piedmont governing body; (b) the so-called Ruby ter trial in which Silvio Berlusconi was accused of corruption; (c) EU Corruption Report; (d) bribery in the Italian Space Agency and news on the EU Corruption Report; (e) appointment of a new prime minister (Matteo Renzi) with statements about the fight against corruption; (f) Maugeri trial in which members of the Lombardy governing body were accused of corruption and embezzlement; (g) corruption discovered in the Lombardy public health system. Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 21 / 24
Comparison between factual and non-factual news (1) (2) (3) (4) (5) (6) Corruption Justice Enforcement Corruption news 0.17* 0.09 0.30*** -0.09-0.06-0.03 (0.09) (0.13) (0.08) (0.06) (0.09) (0.08) Constant 76.77*** 57.76*** 23.57-1.85-6.97 16.67 (17.36) (16.49) (29.32) (7.76) (9.05) (17.83) Adj.R2 0.29 0.30 0.31 0.27 0.28 0.29 N.Clusters 62 57 57 62 57 57 N.Observations 1704 1437 1456 1704 1437 1456 Excluding 2 days with both event-related and non-event-related news Excluding 5 days with main non-event-related news Excluding 5 days with main event-related news Effect driven by non-event-related (non-factual) news persuasion Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 22 / 24
Conclusions: results Increasing individual exposure to corruption news in the day of the interview by one sd (7.5 news) raises corruption perceptions by about 3.5% and reduces trust in justice effectiveness by about 5.2% These are short-term effects The press may not be successful much of the time in telling people what to think, but it is stunningly successful in telling its readers what to think about (Bernard Cohen, 1963) - top-of-the-mind Not all news affect individuals perceptions, their content matters: perceptions react to non-event related news and not to event-related ones Objective facts are less influential in shaping public opinion than appeals to emotion and personal belief - post-truth Good news (investigations, arrests, convictions) do not influence individuals perceptions, whereas bad news (reports, speeches) do - negativity bias Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 23 / 24
Conclusions: policy implications We highlight one potential pitfall of existing corruption measures based on perceptions Incorporate effect of media A short run effect on individuals perceptions may have long run effects on social, political, economic outcomes if news are released on the eve of, for example, elections Caution in the diffusion of certain news Control over possible strategic behaviors on the part of media International organizations have a great responsibility when producing and releasing estimates on sensitive but hardly observable phenomena Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 24 / 24
. Thank you contact: lucia.rizzica@bancaditalia.it Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 25 / 24
Measuring corruption through perceptions The Corruption Perception Index back Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 26 / 24
Data: corruption salience on the media How important are on-line newspapers? Figure: Share of people aged 16-74 who used Internet to read newspapers, news and journals in the previous three months, year 2013. Source: Eurostat 0 20 40 60 UK Germany Spain Greece Portugal France ITALY Belgium back Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 27 / 24
Heterogeneous effects (1) (2) (3) (4) Interactions: College-educated Civil Servant Self-employed Frequent PO user Panel A Corruption Corruption news 0.18** 0.19** 0.15* 0.19* (0.08) (0.08) (0.08) (0.10) Corruption news Interaction -0.17-0.19 0.21-0.05 (0.25) (0.25) (0.31) (0.15) Interaction 0.87-2.61-2.56 3.49 (3.51) (3.97) (3.71) (2.49) Constant 36.76** 37.85** 35.48** 39.23** (16.96) (16.28) (15.87) (16.65) Adj.R2 0.27 0.28 0.28 0.28 Panel C Justice Effectiveness Corruption news -0.08-0.07-0.14** -0.04 (0.06) (0.06) (0.06) (0.07) Corruption news Interaction -0.36* -0.47** 0.33* -0.18** (0.20) (0.23) (0.18) (0.07) Interaction 4.24* 8.01** -5.25** 1.83 (2.52) (3.47) (2.44) (1.36) Constant -18.15** -18.46* -17.25* -13.30 (8.74) (9.29) (8.88) (8.85) Adj.R2 0.26 0.26 0.26 0.26 N.Observations 1805 1805 1805 1805 Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 28 / 24
Persistence of the effects Baseline specification Y idpe = α 0 + βnews dp + β l News (d x)p + α 1 X i + α 2 Int i + ϕ t + ϕ p + ϕ e + ɛ idpe Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 29 / 24
Persistence of the effects Alternative specification Y idpe = α 0 + β 2 News (d x)p }{{} +α 1X i + α 2 Int i + ϕ t + ϕ p + ϕ e + ɛ idpe back Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 30 / 24
Robustness checks Potential threats to identification Variants of the News dp variable: excluding outliers and weighting for each newspaper relative diffusion. Timing of the interview: respondents might be influenced by news in the days before the interview: 3 days MA and a cumulative version of News dp ; we treat differently interviews occurred early in the morning and on week-end days Unobserved heterogeneity: accounting for unobserved factors: province by week FE Endogeneity of the News dp : there are no changes in the Editorial Boards of the newspapers over the period of the interviews sorting of readers across newspapers: controlling for individual attitudes to Tax Evasion (Olken 2009) Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 31 / 24
Robustness checks Panel A (1) (2) (3) (4) (5) (6) (7) (8) Corruption Corruption news 0.18** 0.18** 2.95* 0.15** 0.17** 0.19*** 0.22** 0.16** (0.08) (0.07) (1.63) (0.07) (0.07) (0.07) (0.11) (0.07) Individual attitude towards tax evasion -3.09*** (1.11) Constant 38.56** 55.25** 39.29** 11.14 38.72** 768.10*** 534.15 49.02*** (16.43) (23.59) (16.41) (12.94) (16.45) (287.07) (1015.70) (16.92) Adj.R2 0.28 0.28 0.28 0.27 0.28 0.28 0.31 0.28 Panel C Justice Enforcement Corruption news -0.11* -0.11** -2.65** -0.13** -0.11* -0.08-0.10* -0.11* (0.06) (0.05) (1.14) (0.06) (0.06) (0.06) (0.06) (0.06) Individual attitude towards tax evasion 1.26 (0.80) Constant -16.29* 33.26* -16.33* -2.14-16.41* 309.22 402.00-20.52** (8.95) (18.55) (8.81) (5.55) (8.86) (192.05) (465.03) (9.20) Adj.R2 0.26 0.26 0.26 0.26 0.26 0.26 0.29 0.26 N.Observations 1805 1803 1805 1653 1805 1805 1805 1805 Excluding two outlying newspapers Three days moving average of Newsdp Weighted version of Newsdp Excluding interviews on Saturday and Sunday Morning interviews linked to News (d 1)p Week fixed effects Week by province fixed effects Control for individual attitude towards tax evasion Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 32 / 24
Robustness checks Placebo regression: random match of the dates (A) Corruption Coefficient -.2-.10.1.2 0 200 400 600 800 1000 Replications No. (B) Social Norms Coefficient -.2-.10.1.2 0 200 400 600 800 1000 Replications No. (C) Justice Enforcement Coefficient -.2-.10.1.2 0 200 400 600 800 1000 Replications No. Rizzica, Tonello (Bank of Italy) Persuadable Perceptions 11 May 2017 33 / 24