Newspapers, Local News and Electoral Politics
Francesco DragoUniversity of Napoli, Parthenope; CSEF & IZA
Tommaso NanniciniBocconi University, IGIER & IZA
Francesco SobbrioIMT Lucca
9th Workshop in Media Economics
October 29, 2011
Drago, Nannicini, Sobbrio () Local News and Electoral Politics 1 / 18
Motivations: news media and politics
An active and politically independent news media industry is, usually,considered a vital element for the health of democracies (Thomas Je¤erson,1787; Alexis de Toqueville, 19831; Bagdikian, 2000; Starr, 2009).
News media are supposed/expected to:
I Inform or remind people that an election is taking place.
I Provide information about the issues at stake, the candidates�characteristicsand political platforms.
I Keep politicians accountable.
What is the actual role played by news media on electoral outcomes, selectionof politicians and government�s performance?
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 1 / 18
Motivations: news media and politics
An active and politically independent news media industry is, usually,considered a vital element for the health of democracies (Thomas Je¤erson,1787; Alexis de Toqueville, 19831; Bagdikian, 2000; Starr, 2009).
News media are supposed/expected to:
I Inform or remind people that an election is taking place.
I Provide information about the issues at stake, the candidates�characteristicsand political platforms.
I Keep politicians accountable.
What is the actual role played by news media on electoral outcomes, selectionof politicians and government�s performance?
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 1 / 18
Motivations: news media and politics
An active and politically independent news media industry is, usually,considered a vital element for the health of democracies (Thomas Je¤erson,1787; Alexis de Toqueville, 19831; Bagdikian, 2000; Starr, 2009).
News media are supposed/expected to:
I Inform or remind people that an election is taking place.
I Provide information about the issues at stake, the candidates�characteristicsand political platforms.
I Keep politicians accountable.
What is the actual role played by news media on electoral outcomes, selectionof politicians and government�s performance?
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 1 / 18
Motivations: news media and politics
An active and politically independent news media industry is, usually,considered a vital element for the health of democracies (Thomas Je¤erson,1787; Alexis de Toqueville, 19831; Bagdikian, 2000; Starr, 2009).
News media are supposed/expected to:
I Inform or remind people that an election is taking place.
I Provide information about the issues at stake, the candidates�characteristicsand political platforms.
I Keep politicians accountable.
What is the actual role played by news media on electoral outcomes, selectionof politicians and government�s performance?
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 1 / 18
Motivations: news media and politics
An active and politically independent news media industry is, usually,considered a vital element for the health of democracies (Thomas Je¤erson,1787; Alexis de Toqueville, 19831; Bagdikian, 2000; Starr, 2009).
News media are supposed/expected to:
I Inform or remind people that an election is taking place.
I Provide information about the issues at stake, the candidates�characteristicsand political platforms.
I Keep politicians accountable.
What is the actual role played by news media on electoral outcomes, selectionof politicians and government�s performance?
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 1 / 18
Motivations: news media and politics
An active and politically independent news media industry is, usually,considered a vital element for the health of democracies (Thomas Je¤erson,1787; Alexis de Toqueville, 19831; Bagdikian, 2000; Starr, 2009).
News media are supposed/expected to:
I Inform or remind people that an election is taking place.
I Provide information about the issues at stake, the candidates�characteristicsand political platforms.
I Keep politicians accountable.
What is the actual role played by news media on electoral outcomes, selectionof politicians and government�s performance?
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 1 / 18
This paper: newspapers and political outcomes
1 New dataset covering the presence of local news provided by di¤erent typesof newspapers (provincial, multi-regional, national) for all Italianmunicipalities with more than 15,000 inhabitants in the period 1993-2009.
2 Assess the impact of entry and exit of provincial newspapers or local editionsof multi-regional/national newspapers on municipal politics:
I Electoral outcomes: turnout and incumbent mayor�s reelection.I Selection of politicians: characteristics of elected mayors (i.e., gender andeducation).
I Local governments�performance: taxes, revenues, transfers, e¢ ciencymeasures (not yet).
3 Do di¤erent types of newspapers (provincial, multi-regional, national) havedi¤erent e¤ects on voters�and politicians�characteristics and behavior?
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 2 / 18
This paper: newspapers and political outcomes
1 New dataset covering the presence of local news provided by di¤erent typesof newspapers (provincial, multi-regional, national) for all Italianmunicipalities with more than 15,000 inhabitants in the period 1993-2009.
2 Assess the impact of entry and exit of provincial newspapers or local editionsof multi-regional/national newspapers on municipal politics:
I Electoral outcomes: turnout and incumbent mayor�s reelection.I Selection of politicians: characteristics of elected mayors (i.e., gender andeducation).
I Local governments�performance: taxes, revenues, transfers, e¢ ciencymeasures (not yet).
3 Do di¤erent types of newspapers (provincial, multi-regional, national) havedi¤erent e¤ects on voters�and politicians�characteristics and behavior?
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 2 / 18
This paper: newspapers and political outcomes
1 New dataset covering the presence of local news provided by di¤erent typesof newspapers (provincial, multi-regional, national) for all Italianmunicipalities with more than 15,000 inhabitants in the period 1993-2009.
2 Assess the impact of entry and exit of provincial newspapers or local editionsof multi-regional/national newspapers on municipal politics:
I Electoral outcomes: turnout and incumbent mayor�s reelection.
I Selection of politicians: characteristics of elected mayors (i.e., gender andeducation).
I Local governments�performance: taxes, revenues, transfers, e¢ ciencymeasures (not yet).
3 Do di¤erent types of newspapers (provincial, multi-regional, national) havedi¤erent e¤ects on voters�and politicians�characteristics and behavior?
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 2 / 18
This paper: newspapers and political outcomes
1 New dataset covering the presence of local news provided by di¤erent typesof newspapers (provincial, multi-regional, national) for all Italianmunicipalities with more than 15,000 inhabitants in the period 1993-2009.
2 Assess the impact of entry and exit of provincial newspapers or local editionsof multi-regional/national newspapers on municipal politics:
I Electoral outcomes: turnout and incumbent mayor�s reelection.I Selection of politicians: characteristics of elected mayors (i.e., gender andeducation).
I Local governments�performance: taxes, revenues, transfers, e¢ ciencymeasures (not yet).
3 Do di¤erent types of newspapers (provincial, multi-regional, national) havedi¤erent e¤ects on voters�and politicians�characteristics and behavior?
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 2 / 18
This paper: newspapers and political outcomes
1 New dataset covering the presence of local news provided by di¤erent typesof newspapers (provincial, multi-regional, national) for all Italianmunicipalities with more than 15,000 inhabitants in the period 1993-2009.
2 Assess the impact of entry and exit of provincial newspapers or local editionsof multi-regional/national newspapers on municipal politics:
I Electoral outcomes: turnout and incumbent mayor�s reelection.I Selection of politicians: characteristics of elected mayors (i.e., gender andeducation).
I Local governments�performance: taxes, revenues, transfers, e¢ ciencymeasures (not yet).
3 Do di¤erent types of newspapers (provincial, multi-regional, national) havedi¤erent e¤ects on voters�and politicians�characteristics and behavior?
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 2 / 18
This paper: newspapers and political outcomes
1 New dataset covering the presence of local news provided by di¤erent typesof newspapers (provincial, multi-regional, national) for all Italianmunicipalities with more than 15,000 inhabitants in the period 1993-2009.
2 Assess the impact of entry and exit of provincial newspapers or local editionsof multi-regional/national newspapers on municipal politics:
I Electoral outcomes: turnout and incumbent mayor�s reelection.I Selection of politicians: characteristics of elected mayors (i.e., gender andeducation).
I Local governments�performance: taxes, revenues, transfers, e¢ ciencymeasures (not yet).
3 Do di¤erent types of newspapers (provincial, multi-regional, national) havedi¤erent e¤ects on voters�and politicians�characteristics and behavior?
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 2 / 18
Overview of main results1 Di¤erent types of newspapers have di¤erent e¤ects on electoral participation.
I An additional provincial newspaper increases turnout in municipal elections by1%
I An additional multi-regional newspaper increases turnout in municipalelections by 5.3%
I The presence of a local edition of a national newspapers does not have anysigni�cant e¤ect (partial crowding-out)
2 The entry and exit of newspapers do not a¤ect the probability of incumbentmayor being reelected.
3 Conditional on (2), newspapers have a signi�cant e¤ect on the selection ofpoliticians. Speci�cally, an additional multi-regional newspaper supplyinglocal news:
I Increases the probability of a �freshman�mayor being a woman by 14percentage points (underdog e¤ect)
I Increases the probability of a �freshman�mayor having a college degree by 27percentage points (quality e¤ect).
Related literature
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 3 / 18
Overview of main results1 Di¤erent types of newspapers have di¤erent e¤ects on electoral participation.
I An additional provincial newspaper increases turnout in municipal elections by1%
I An additional multi-regional newspaper increases turnout in municipalelections by 5.3%
I The presence of a local edition of a national newspapers does not have anysigni�cant e¤ect (partial crowding-out)
2 The entry and exit of newspapers do not a¤ect the probability of incumbentmayor being reelected.
3 Conditional on (2), newspapers have a signi�cant e¤ect on the selection ofpoliticians. Speci�cally, an additional multi-regional newspaper supplyinglocal news:
I Increases the probability of a �freshman�mayor being a woman by 14percentage points (underdog e¤ect)
I Increases the probability of a �freshman�mayor having a college degree by 27percentage points (quality e¤ect).
Related literature
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 3 / 18
Overview of main results1 Di¤erent types of newspapers have di¤erent e¤ects on electoral participation.
I An additional provincial newspaper increases turnout in municipal elections by1%
I An additional multi-regional newspaper increases turnout in municipalelections by 5.3%
I The presence of a local edition of a national newspapers does not have anysigni�cant e¤ect (partial crowding-out)
2 The entry and exit of newspapers do not a¤ect the probability of incumbentmayor being reelected.
3 Conditional on (2), newspapers have a signi�cant e¤ect on the selection ofpoliticians. Speci�cally, an additional multi-regional newspaper supplyinglocal news:
I Increases the probability of a �freshman�mayor being a woman by 14percentage points (underdog e¤ect)
I Increases the probability of a �freshman�mayor having a college degree by 27percentage points (quality e¤ect).
Related literature
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 3 / 18
Overview of main results1 Di¤erent types of newspapers have di¤erent e¤ects on electoral participation.
I An additional provincial newspaper increases turnout in municipal elections by1%
I An additional multi-regional newspaper increases turnout in municipalelections by 5.3%
I The presence of a local edition of a national newspapers does not have anysigni�cant e¤ect (partial crowding-out)
2 The entry and exit of newspapers do not a¤ect the probability of incumbentmayor being reelected.
3 Conditional on (2), newspapers have a signi�cant e¤ect on the selection ofpoliticians. Speci�cally, an additional multi-regional newspaper supplyinglocal news:
I Increases the probability of a �freshman�mayor being a woman by 14percentage points (underdog e¤ect)
I Increases the probability of a �freshman�mayor having a college degree by 27percentage points (quality e¤ect).
Related literature
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 3 / 18
Overview of main results1 Di¤erent types of newspapers have di¤erent e¤ects on electoral participation.
I An additional provincial newspaper increases turnout in municipal elections by1%
I An additional multi-regional newspaper increases turnout in municipalelections by 5.3%
I The presence of a local edition of a national newspapers does not have anysigni�cant e¤ect (partial crowding-out)
2 The entry and exit of newspapers do not a¤ect the probability of incumbentmayor being reelected.
3 Conditional on (2), newspapers have a signi�cant e¤ect on the selection ofpoliticians. Speci�cally, an additional multi-regional newspaper supplyinglocal news:
I Increases the probability of a �freshman�mayor being a woman by 14percentage points (underdog e¤ect)
I Increases the probability of a �freshman�mayor having a college degree by 27percentage points (quality e¤ect).
Related literature
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 3 / 18
Overview of main results1 Di¤erent types of newspapers have di¤erent e¤ects on electoral participation.
I An additional provincial newspaper increases turnout in municipal elections by1%
I An additional multi-regional newspaper increases turnout in municipalelections by 5.3%
I The presence of a local edition of a national newspapers does not have anysigni�cant e¤ect (partial crowding-out)
2 The entry and exit of newspapers do not a¤ect the probability of incumbentmayor being reelected.
3 Conditional on (2), newspapers have a signi�cant e¤ect on the selection ofpoliticians. Speci�cally, an additional multi-regional newspaper supplyinglocal news:
I Increases the probability of a �freshman�mayor being a woman by 14percentage points (underdog e¤ect)
I Increases the probability of a �freshman�mayor having a college degree by 27percentage points (quality e¤ect).
Related literature
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 3 / 18
Overview of main results1 Di¤erent types of newspapers have di¤erent e¤ects on electoral participation.
I An additional provincial newspaper increases turnout in municipal elections by1%
I An additional multi-regional newspaper increases turnout in municipalelections by 5.3%
I The presence of a local edition of a national newspapers does not have anysigni�cant e¤ect (partial crowding-out)
2 The entry and exit of newspapers do not a¤ect the probability of incumbentmayor being reelected.
3 Conditional on (2), newspapers have a signi�cant e¤ect on the selection ofpoliticians. Speci�cally, an additional multi-regional newspaper supplyinglocal news:
I Increases the probability of a �freshman�mayor being a woman by 14percentage points (underdog e¤ect)
I Increases the probability of a �freshman�mayor having a college degree by 27percentage points (quality e¤ect).
Related literature
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 3 / 18
Overview of main results1 Di¤erent types of newspapers have di¤erent e¤ects on electoral participation.
I An additional provincial newspaper increases turnout in municipal elections by1%
I An additional multi-regional newspaper increases turnout in municipalelections by 5.3%
I The presence of a local edition of a national newspapers does not have anysigni�cant e¤ect (partial crowding-out)
2 The entry and exit of newspapers do not a¤ect the probability of incumbentmayor being reelected.
3 Conditional on (2), newspapers have a signi�cant e¤ect on the selection ofpoliticians. Speci�cally, an additional multi-regional newspaper supplyinglocal news:
I Increases the probability of a �freshman�mayor being a woman by 14percentage points (underdog e¤ect)
I Increases the probability of a �freshman�mayor having a college degree by 27percentage points (quality e¤ect).
Related literature
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 3 / 18
Background. The Italian Newspaper Industry.
1 Geographical segmentation of news media in Italy: 93% of the Italian
television market dominated by national TV stations (AGCOM, 2009) !Newspapers still remain the main, if not unique, source of informationat the local level (Sorrentino, 2004; Murialdi and Tranfaglia, 2008).
2 Technological innovations: from the mid-80s, newspapers abandoned the
linotype and other �hot-metal� typesetting machines by adoptingphotosetting and, later on, computerized typesetting =) Decrease in thecost and the length of the production process. Expansion of local news
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 4 / 18
Background. The Italian Newspaper Industry.
1 Geographical segmentation of news media in Italy: 93% of the Italian
television market dominated by national TV stations (AGCOM, 2009) !Newspapers still remain the main, if not unique, source of informationat the local level (Sorrentino, 2004; Murialdi and Tranfaglia, 2008).
2 Technological innovations: from the mid-80s, newspapers abandoned the
linotype and other �hot-metal� typesetting machines by adoptingphotosetting and, later on, computerized typesetting =) Decrease in thecost and the length of the production process. Expansion of local news
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 4 / 18
Background. Italian municipalities.
Since March 1993, mayors are directly elected with plurality rule and aresubject to a two-term limit.
Italian municipalities are in charge of a wide range of services (e.g., watersupply, waste management, municipal police, etc.).
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 5 / 18
Data, market de�nition and sample selection
For each year between 1993 and 2009 ! We identify, for each municipalitywith more than 15,000 people, the number of provincial newspapers/localedition of multi-regional and national newspapers providing news on thatmunicipality. Data Newspaper Panel
Both our newspaper data and electoral data are at the city-level =) wede�ne the news market to be a city =) perfect overlap between the newsmarket and the outcomes of interest.
Electoral/Newspapers data on 758 Italian cities between 1993 and2009.
542 city-years with net newspapers entry and 137 city-years with netnewspapers exit.
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 6 / 18
Data, market de�nition and sample selection
For each year between 1993 and 2009 ! We identify, for each municipalitywith more than 15,000 people, the number of provincial newspapers/localedition of multi-regional and national newspapers providing news on thatmunicipality. Data Newspaper Panel
Both our newspaper data and electoral data are at the city-level =) wede�ne the news market to be a city =) perfect overlap between the newsmarket and the outcomes of interest.
Electoral/Newspapers data on 758 Italian cities between 1993 and2009.
542 city-years with net newspapers entry and 137 city-years with netnewspapers exit.
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 6 / 18
Data, market de�nition and sample selection
For each year between 1993 and 2009 ! We identify, for each municipalitywith more than 15,000 people, the number of provincial newspapers/localedition of multi-regional and national newspapers providing news on thatmunicipality. Data Newspaper Panel
Both our newspaper data and electoral data are at the city-level =) wede�ne the news market to be a city =) perfect overlap between the newsmarket and the outcomes of interest.
Electoral/Newspapers data on 758 Italian cities between 1993 and2009.
542 city-years with net newspapers entry and 137 city-years with netnewspapers exit.
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 6 / 18
Data, market de�nition and sample selection
For each year between 1993 and 2009 ! We identify, for each municipalitywith more than 15,000 people, the number of provincial newspapers/localedition of multi-regional and national newspapers providing news on thatmunicipality. Data Newspaper Panel
Both our newspaper data and electoral data are at the city-level =) wede�ne the news market to be a city =) perfect overlap between the newsmarket and the outcomes of interest.
Electoral/Newspapers data on 758 Italian cities between 1993 and2009.
542 city-years with net newspapers entry and 137 city-years with netnewspapers exit.
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 6 / 18
Empirical framework
We estimate the e¤ect of newspapers on political outcomes by using followingregression model:
yit = θht + γi + β � nit +~α �~xit + λ � zit + εit , (1)
i denotes the city, h the macro-region and t the election year.
Fixed e¤ects: θht macro-region-by-year �xed e¤ects, γi city-�xed e¤ects.
nit : number of newspapers providing local news on city i in year t
~α is a vector of parameters and ~xit is a vector of control variables at the cityand/or province level changing over time (i.e., population, unemploymentrate and di¤erence between new and ceased �rms in year t).
zit : unobserved pro�tability of local news in city i in year t ! λ is the degreeof correlation between pro�tability and the electoral outcome once we controlfor ~xit , nit and the �xed e¤ects θht and γi .
εit :idiosyncratic unobserved error
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 7 / 18
Empirical Strategy.
Investigate how a (discrete) increase or decrease in the supply of newspapersproviding news on a given municipality a¤ects political outcomes in thatmunicipality, relative to other municipalities in the same macro-region andyear that do not experience it.
Possible issues: the potential correlation between unobserved pro�tability zitand yit may bias our results.
! The pro�tability of local news in city i at time t, i.e., zit , maybecorrelated to electoral outcomes (e.g., turnout) even after controlling for�xed e¤ects and observable characteristics, i.e., λ 6= 0 ! bias arising fromcorrelation between zit and nit .
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 8 / 18
Empirical Strategy.
Investigate how a (discrete) increase or decrease in the supply of newspapersproviding news on a given municipality a¤ects political outcomes in thatmunicipality, relative to other municipalities in the same macro-region andyear that do not experience it.
Possible issues: the potential correlation between unobserved pro�tability zitand yit may bias our results.
! The pro�tability of local news in city i at time t, i.e., zit , maybecorrelated to electoral outcomes (e.g., turnout) even after controlling for�xed e¤ects and observable characteristics, i.e., λ 6= 0 ! bias arising fromcorrelation between zit and nit .
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 8 / 18
Identi�cation of causal e¤ects. Discrete changes in thenumber of newspapers
The identi�cation exploits discrete changes in the number of newspapers indi¤erent election years.
The entry of a newspaper arguably depends on a thresholds value z� !whenever zit > z�, we should observe a newspaper entry.
Discrete changes in the number of newspapers ! entrance of a newspaperdepends on whether zit is above this threshold z� rather than on the distanceof zit from it.
Periods with large variations in zit and no entry (i.e., zit remains below z�)and periods with small variations in zit and entry (i.e., zit goes above z�).(Gentzkow et al. 2010).
I Even if λ 6= 0, the variation in pro�tability between t � 1 and t explains asmall part of the variation in nit .
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 9 / 18
Identi�cation of causal e¤ects. Discrete changes in thenumber of newspapers
The identi�cation exploits discrete changes in the number of newspapers indi¤erent election years.
The entry of a newspaper arguably depends on a thresholds value z� !whenever zit > z�, we should observe a newspaper entry.
Discrete changes in the number of newspapers ! entrance of a newspaperdepends on whether zit is above this threshold z� rather than on the distanceof zit from it.
Periods with large variations in zit and no entry (i.e., zit remains below z�)and periods with small variations in zit and entry (i.e., zit goes above z�).(Gentzkow et al. 2010).
I Even if λ 6= 0, the variation in pro�tability between t � 1 and t explains asmall part of the variation in nit .
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 9 / 18
Identi�cation of causal e¤ects. Discrete changes in thenumber of newspapers
The identi�cation exploits discrete changes in the number of newspapers indi¤erent election years.
The entry of a newspaper arguably depends on a thresholds value z� !whenever zit > z�, we should observe a newspaper entry.
Discrete changes in the number of newspapers ! entrance of a newspaperdepends on whether zit is above this threshold z� rather than on the distanceof zit from it.
Periods with large variations in zit and no entry (i.e., zit remains below z�)and periods with small variations in zit and entry (i.e., zit goes above z�).(Gentzkow et al. 2010).
I Even if λ 6= 0, the variation in pro�tability between t � 1 and t explains asmall part of the variation in nit .
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 9 / 18
Identi�cation of causal e¤ects. No correlation with pastchanges in political outcomes
Number ofprovincial
newspapers
Number ofmultiregionalnewspapers
Number ofnational
newspapersTurnout (t1) .7483 .1797 .7259
(1.489) (.3435) (.6773)Turnout (t2) 2.387* .2343 .1102
( 1.327) (.3206) (.3218)
Ftest, both coefficients=0 1.74 0.27 0.85
City and provincial characteristics YES YES YES
City fixed effects YES YES YES
Macroregionalbyyear fixed effects YES YES YES
Rsquared 0.936 0.984 0.904Number of cities 610 610 610Observations 1038 1038 1038
Previous turnout rates and newspapers
Note: Standard errors in parentheses are clustered at the province level.
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 10 / 18
Identi�cation of causal e¤ects. Technological innovations.
Overall, the evolution of the Italian newspaper industry suggests thattechnological innovations have been, most likely, the key drivers of theobserved expansion in the supply of local news.
Negative shocks in the newspapers�cost function have decreased the grossrevenues threshold above which newspapers �nd optimal to enter in a localmarket ! indirectly induced the entry of additional newspapers in the localmarket for news.
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 11 / 18
Turnout Turnout Turnout
.0083** (.0043)
.0431*** (.0081)
.0038(.0056)
City and provincial characteristics YES YES YES
City fixed effects YES YES YES
Macroregionalbyyear fixed effects YES YES YES
Average dependent variable 0.792 0.792 0.792Rsquared 0.903 0.904 0.902Number of cities 698 698 698Observations 2365 2365 2365Note: Standard errors in parentheses are clustered at the province level.
Newspapers and turnout
Number of provincial newspapers
Number of multiregional newspapers
Number of national newspapers
Theory and previous evidence
Newspapers competition and turnout
Discussion and magnitudes
An additional multi-regional (provincial) newspaper increases turnout 5.3%(1% )
Partial crowding-out e¤ect on local participation by former readers of thelocal incumbent newspaper, created by the entrance of a national newspaperin a local market for news (George and Waldfogel, 2003; 2008).
Magnitudes of these e¤ects in line with most of the literature (Strömberg,2004; Gentzkow, 2006; Oberholzer-Gee and Waldfogel, 2010) butsubstantially larger than the one found by Gentzkow et al. (2010), i.e., 0,3percentage points.
Similarly to Oberholzer-Gee and Waldfogel (2010) and unlike Gentzkow et al.(2010), we estimate the e¤ect of local news on local elections ! theinformational content of local news provided by a newspaper in our sample islikely to be very salient.
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 13 / 18
Discussion and magnitudes
An additional multi-regional (provincial) newspaper increases turnout 5.3%(1% )
Partial crowding-out e¤ect on local participation by former readers of thelocal incumbent newspaper, created by the entrance of a national newspaperin a local market for news (George and Waldfogel, 2003; 2008).
Magnitudes of these e¤ects in line with most of the literature (Strömberg,2004; Gentzkow, 2006; Oberholzer-Gee and Waldfogel, 2010) butsubstantially larger than the one found by Gentzkow et al. (2010), i.e., 0,3percentage points.
Similarly to Oberholzer-Gee and Waldfogel (2010) and unlike Gentzkow et al.(2010), we estimate the e¤ect of local news on local elections ! theinformational content of local news provided by a newspaper in our sample islikely to be very salient.
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 13 / 18
Probability ofincumbent mayor
reelected
Probability ofincumbent mayor
reelected
Probability ofincumbent mayor
reelected
.0272 (.0477)
.0153 (.1894)
.0386(.0668)
City and provincial characteristics YES YES YES
City fixed effects YES YES YES
Macroregionalbyyear fixed effects YES YES YES
Average dependent variable 0.258 0.258 0.258Rsquared 0.362 0.361 0.366Number of cities 690 690 690Observations 2261 2261 2261
Newspapers and incumbent's reelection
Note: Standard errors in parentheses are clustered at the province level.
Number of provincial newspapers
Number of multiregional newspapers
Number of national newspapers
Theory and previous evidence
Gender Years ofSchooling College
.0356 .1771 .0418(.0234) (.1970) ( .0397) .0459 1.021 .0473(.0464) ( 1.194) (.1799).0239 .5077 .0808
(.0651) (.3662) (.0778)
City and provincial characteristics YES YES YES
City fixed effects YES YES YES
Macroregionalbyyear fixed effects YES YES YES
Average dependent variable 0.069 15.38 0.646Number of cities 690 687 687Observations 2261 2223 2223Note: Standard errors in parentheses are clustered at the province level.
Newspapers and selection of mayors
Number of provincial newspapers
Number of multiregional newspapers
Number of national newspapers
Gender conditional on
freshman mayor
Years of Schooling conditional on
freshman mayor
College conditional on
freshman mayor
.0522 .1963 .0576(.0345) ( .2438) ( .0519).1466* 1.401* .2762*( .0741) ( .8222) (.1637).0252 .6013 .0980
(.0252) (.4043) (.0923)
City and provincial characteristics YES YES YES
City fixed effects YES YES YES
Macroregionalbyyear fixed effects YES YES YES
Average dependent variable 0.073 15.45 0.661Number of cities 684 680 680Observations 1678 1645 1645Note: Standard errors in parentheses are clustered at the province level.
Newspapers and selection of freshmen mayors
Number of provincial newspapers
Number of multiregional newspapers
Number of national newspapers
Discussion: newspapers and selection of politicians
Two main e¤ects of an increase in the supply of newspapers on the selection ofpoliticians.
1 Sizeable positive e¤ect of an increase in the supply of newspapers providinglocal news on the education of elected mayors (quality e¤ect)
2 Lowering barriers to entry into political representation for candidates that areex-ante less favored (percentage of women political representation in Italy isextremely low) (underdog e¤ect)
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 16 / 18
Discussion: newspapers and selection of politicians
Two main e¤ects of an increase in the supply of newspapers on the selection ofpoliticians.
1 Sizeable positive e¤ect of an increase in the supply of newspapers providinglocal news on the education of elected mayors (quality e¤ect)
2 Lowering barriers to entry into political representation for candidates that areex-ante less favored (percentage of women political representation in Italy isextremely low) (underdog e¤ect)
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 16 / 18
Conclusions
Preliminary evidence seems to suggest that provincial and multi-regionalnewspapers have a signi�cant e¤ect on electoral participation (1% to 5.3%).No signi�cant e¤ect of national newspapers (partial crowding-out).
No e¤ect of newspapers on incumbents�reelection probability.
Selection of �freshmen�mayors consistent with the patterns found onelectoral participation. An additional multi-regional newspaper supplyinglocal news:
I Increases the probability of a �freshman�mayor being a woman by 14percentage points (underdog e¤ect)
I Increases the probability of a �freshman�mayor having a college degree by 27percentage points (quality e¤ect)
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 17 / 18
Conclusions
Preliminary evidence seems to suggest that provincial and multi-regionalnewspapers have a signi�cant e¤ect on electoral participation (1% to 5.3%).No signi�cant e¤ect of national newspapers (partial crowding-out).
No e¤ect of newspapers on incumbents�reelection probability.
Selection of �freshmen�mayors consistent with the patterns found onelectoral participation. An additional multi-regional newspaper supplyinglocal news:
I Increases the probability of a �freshman�mayor being a woman by 14percentage points (underdog e¤ect)
I Increases the probability of a �freshman�mayor having a college degree by 27percentage points (quality e¤ect)
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 17 / 18
Conclusions
Preliminary evidence seems to suggest that provincial and multi-regionalnewspapers have a signi�cant e¤ect on electoral participation (1% to 5.3%).No signi�cant e¤ect of national newspapers (partial crowding-out).
No e¤ect of newspapers on incumbents�reelection probability.
Selection of �freshmen�mayors consistent with the patterns found onelectoral participation. An additional multi-regional newspaper supplyinglocal news:
I Increases the probability of a �freshman�mayor being a woman by 14percentage points (underdog e¤ect)
I Increases the probability of a �freshman�mayor having a college degree by 27percentage points (quality e¤ect)
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 17 / 18
Conclusions
Preliminary evidence seems to suggest that provincial and multi-regionalnewspapers have a signi�cant e¤ect on electoral participation (1% to 5.3%).No signi�cant e¤ect of national newspapers (partial crowding-out).
No e¤ect of newspapers on incumbents�reelection probability.
Selection of �freshmen�mayors consistent with the patterns found onelectoral participation. An additional multi-regional newspaper supplyinglocal news:
I Increases the probability of a �freshman�mayor being a woman by 14percentage points (underdog e¤ect)
I Increases the probability of a �freshman�mayor having a college degree by 27percentage points (quality e¤ect)
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 17 / 18
Conclusions
Preliminary evidence seems to suggest that provincial and multi-regionalnewspapers have a signi�cant e¤ect on electoral participation (1% to 5.3%).No signi�cant e¤ect of national newspapers (partial crowding-out).
No e¤ect of newspapers on incumbents�reelection probability.
Selection of �freshmen�mayors consistent with the patterns found onelectoral participation. An additional multi-regional newspaper supplyinglocal news:
I Increases the probability of a �freshman�mayor being a woman by 14percentage points (underdog e¤ect)
I Increases the probability of a �freshman�mayor having a college degree by 27percentage points (quality e¤ect)
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 17 / 18
Next
Expand electoral and newspapers dataset (analyze also impact of newspaperson national elections).
Analyze incumbency advantage (rather than just prob. of reelection)
Data on local governments�outcomes (revenues, taxes, de�cit, transfers,e¢ ciency measures) ! Assess the e¤ect of newspapers on politicians�e¤ort.
Drago, Nannicini, Sobbrio (October, 29, 2011) Local News and Electoral Politics 18 / 18
Related literature.
1 News media and electoral outcomes (turnout and incumbency advantage):
Strömberg, 2004, QJE ; Gentzkow, 2006, QJE ; Ferraz and Finan, 2008, QJE ;George and Waldfogel, 2008, Gentzkow et al., 2010, AER; Oberholzer-Geeand Waldfogel, 2010, AER; Snyder and Strömberg, 2010, JPE ; Cagé andKeslair, 2011; Schulhofer-Wohl and Garrido, 2011.
2 News media and selection of politicians: Snyder-Strömberg, 2010, JPE.back
News media and turnout
1 Theory: Information and turnout: Matsusaka (1995, PubChoice), Feddersenand Pesendorfer (1996, AER; 1999, APSR), more informed individuals aremore likely to turnout.
2 Evidence:
1 Informed individuals are more likely to turn-out (Lassen, 2005, AJPS ; Deganand Merlo, 2009, JEEA; Green and Gerber, 2008; Larcinese, 2009, JTPol ;Sobbrio and Navarra, 2010, EJPE ).
2 Positive e¤ects of news media on turnout (Strömberg, 2004, QJE ; Gentzkow,2006, QJE ; Gentzkow et al., 2010, AER ; Oberholzer-Gee and Waldfogel, 2010,AER ; Snyder and Strömberg, 2010; JPE, Schulhofer-Wohl and Garrido, 2011)
3 Negative e¤ects of news media on turnout (Gentzkow, 2006, QJE ; Cagè andKeslair, 2011).
back
News media and incumbent�s reelectionSnyder and Strömberg (2010, JPE ) �It is not obvious that voters with moreinformation will be more supportive of incumbents.�
1 Non univocal theoretical prediction:
1 Besley and Prat (2006, AER): increase in the number of media outletsdecreases the probability of media capture by the incumbent ! decreases theincumbency advantage and increases the turnover of politicians.
2 Prat and Strömberg (2011): increase in the amount of media coverageincreases, on average, the incumbent�s vote share.
2 Mixed evidence:
1 Snyder and Strömberg (2010, JPE ): small e¤ect of an increase in mediacongruence on the incumbency advantage of US House representatives.
2 Prior (2006, JPolitics): weakly signi�cant e¤ect of the number of local TVstations on the incumbency advantage between 1960 and 1970 but not before1960.
3 Ansolabehere, Snowberg and Snyder (2006, Legislative Studies Quart.): donot �nd any signi�cant e¤ect of television coverage on incumbency advantage.
4 Gentzkow et al. (2010, AER) do not �nd any clear evidence of newspapersincreasing or decreasing the incumbency advantage.
back
Technological innovations & expansion of local news
Decrease in �xed cost of production. Even small production facilities becomepro�table (Grandinetti, 2008; Isnenghi, 2008) ! increase in the number ofprovincial newspapers
0
10
20
30
40
50
60
70
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
n_provincial
back
Technological innovations & expansion of local news (2)
Increase in Synergies. Sharp decrease in the cost of exchanging news agencyreleases, ads, materials among connected newspapers ! expansion in the supplyof local news provided by national and multi-regional newspapers
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
0.42
0.44
0.46
0.48
0.5
0.52
0.54
0.56
mean_national mean_multiregional
back
Expansion of local news
Number of newspapers 1985 Changes in number of newspapers 19852009
(2,7](1,2][0,1]
(2,4](1,2][1,1]
back
Expansion of local news
Newspapers per 10,000 people, 1985 Changes in number of newspapers per 10,000 people, 19852009
(.28112,.6961364](.1560093,.28112](.0977823,.1560093][0,.0977823]
(.2247577,.8960573](.115154,.2247577](.0286016,.115154][.2321964,.0286016]
back
The Italian Newspaper Industry. Circulation.
The pattern of circulation over time does not seem to suggest any positive trendin readership levels (i.e., no endogenous increase in the willingness to becomeinformed about political news by Italian citizens).
0100000200000300000400000500000600000700000800000900000
1000000
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
0
500000
1000000
1500000
2000000
2500000
3000000
provincial multiregional national
back
Identi�cation of causal e¤ects: timing of newspaperentrance/exit
More than 80% of net entries of newspapers occurs in years with nomunicipal elections.
Since elections take place every 4/5 years this is the fraction one wouldexpect if entries and exits were as good as random (with respect to thetiming of elections).
Identi�cation issues (2): media capture
Incumbent politicians may manipulate the supply of newspapers providinglocal news in order to in�uence their reelection probability.
Politicians control over media is a serious concern at the national level inItaly. However, this does not seem to be the case at local level:
1 A large fraction of newspapers in our sample provide local news (also) on citiesdi¤erent from the one where they are headquartered.
2 Number of newspapers providing local news is not signi�catively correlatedwith incumbent�s reelection probability.
Turnout Turnout TurnoutCity has 0 provincial newspaper at t1 .0087**
(.0046)City has 1 or more provincial newspaper at t1 .0086
(.005)City has 0 multiregional newspaper at t1 .0386***
(.0077)City has 1 or more multiregional newspaper at t1 .0170
(.0116)City has 0 national newspaper at t1 .0045
(.0076)City has 1 or more national newspaper at t1 .0082
( .0067)
City and provincial characteristics YES YES YES
City fixed effects YES YES YES
Macroregionalbyyear fixed effects YES YES YES
Rsquared 0.901 0.901 0.899Number of cities 655 655 655Observations 1693 1693 1693
Newspapers competition and turnout
Note: Standard errors in parentheses are clustered at the province level
back
Data. Newspapers panel
Directory of newspapers from annual report of the Italian Authority forCommunication (AGCOM) 1985-2009 (name, location, owners, etc.).
Multi-regional and national newspapers: info on the presence of local editionsby cross-checking a variety of sources (i.e., directly contacting the newspapers;Iannace, 1989; Lenzi, 2001; Grandinetti, 2008; �Il Grande Libro della StampaItaliana�, 1993-2010).
For each year ! identify whether a provincial newspaper/local edition of amulti-regional or national newspaper, provided only news on the capital ofthe province (capoluogo di provincia) or it also provided news on the cities ofthat province.
Impute the presence of local news provided by newspaper k for any city iwith more than 15,000 people belonging to province j whenever newspaper kwas providing local news on province j .
back
Data. Electoral outcomes and politicians�characteristics
Electoral Data from the Italian Minister for internal a¤airs.
Data on the individual characteristics of elected politicians from the �PublicAdministrator Registry� (Anagrafe degli Amministratori) by the Minister forInternal A¤airs. (�rst and last name of each elected politician, level ofeducation, profession, date of birth, gender, political a¢ liation, date ofappointment).
Motivations: struggling newspaper industry
Change in per cent in titles versus per cent change in paid circulation, OECD,2000-2008
Public subsidies to the pressDirect subsidies in OECD countries in 2008:
Italy: every year, around 500 million euros are spent in public subsidies (direct andindirect) to the newspaper industry ! not assigned according to criteria basedon pluralism or e¢ ciency but exogenously targeted to speci�c newspapers�categories (e.g., newspapers owned by political parties or by religiousorganizations).
Market structure transition matrix
Table 1: Market structure transition matrix0 1 2 3 4 5+
0 Newspapers 67 4 0 0 01 Newspaper 2 116 22 0 0
2 Newspapers 0 4 110 4 03 Newspapers 0 2 23 30 04 Newspapers 0 0 0 5 9
5+ Newspapers 0 0 0 0 1Notes: Table shows number of cityyears in sample experiencing a given transition