"The Production of Information in an Online World" with Nicolas Hervé (INA) and Marie-Luce Viaud (INA), Review of Economic Studies, 2020, 87(5): 2126-2164.
News production requires investment, and competitors’ ability to appropriate a story may reduce a media’s incentives to provide original content. Yet, there is little legal protection of intellectual property rights in online news production, which raises the issue of the extent of copying online and the incentives to provide original content. In this paper, we build a unique dataset combining all the online content produced by French news media during the year 2013 with new micro audience data. We develop a topic detection algorithm that identifies each news event, trace the timeline of each story, and study news propagation. We provide new evidence on online news production. First, we document high reactivity of online media: one quarter of the news stories are reproduced online in under 4 minutes. We show that this is accompanied by substantial copying, both at the extensive and at the intensive margins, which may constitute a severe threat to the commercial viability of the news media. Next, we estimate the returns to originality in online news production. Using article-level variations and media-level daily audience combined with article-level social media statistics, we find that original content producers tend to receive more viewers, thereby mitigating the newsgathering incentive problem raised by copying.
“Media Competition, Information Provision and Political Participation: Evidence from French Local Newspapers and Elections, 1944-2014”, Journal of Public Economics, 2020, 85.
This paper investigates the impact of increased media competition on the quantity and quality of news provided and, ultimately, on political participation. I build a new county-level panel dataset of local newspaper presence, newspapers’ number of journalists, costs and revenues and political turnout in France, from 1944 to 2014. I estimate the effect of newspaper entry by comparing counties that experience entry to similar counties in the same years that do not. Both sets of counties exhibit similar trends prior to entry, but those with entry experience substantial declines in the average number of journalists. An increased number of newspapers is also associated with fewer articles and less hard news provision. Newspaper entry, and the associated decline in information provision, is ultimately found to decrease voter turnout at local elections. Exploiting the long time span covered by my data, I discuss a number of mechanisms that may drive these empirical findings. First, I examine the relationship between increased competition and media capture in the aftermath of WW2, when newspapers were biased and the advertising market was underdeveloped. I then show that in the recent period the effects are stronger in counties with more homogeneous populations, as predicted by a vertical product differentiation framework, whereas there is little impact in counties with more heterogeneous populations.
"Sex and the Mission: The Conflicting Effects of Early Christian Investments on sub-Saharan Africa's HIV Epidemic" with Valeria Rueda, Journal of Demographic Economics, 2020, 86(3): 213-257.
This article investigates the long-term impact of historical missionary activity on HIV in sub-Saharan Africa. On the one hand, missionaries were the first to invest in modern medicine in the region. On the other hand, Christianity influenced sexual beliefs and behaviors that affect the risk of contagion. We build a new geocoded dataset locating Protestant and Catholic missions in the early 20th century, as well as the health facilities they invested in, that we combine with individual-level DHS data. With these data, we can address separately these two channels, within regions close to historical missionary settlements. First, we show that proximity to historical missionary health facilities decreases the likelihood of HIV; persistence in healthcare provision and safer sexual behaviors in the region explain this result. Second, we show that regions close to historical missionary settlements exhibit higher likelihood of HIV. This effect is driven by the Christian population in our sample. This suggests conversion to Christianity as a possible explanatory channel. Our findings are robust to alternative specifications addressing selection.
"Newspapers in Times of Low Advertising Revenues" with Charles Angelucci, American Economic Journal: Microeconomics, 2019, 11(3): 319-364.
We model the consequences on newspapers’ content and prices of a reduction in advertising revenues. Newspapers choose the size of their newsroom, and readers are heterogeneous in their ideal amount of journalistic-intensive content. We show that a reduction in advertising revenues lowers newspapers’ incentives to pro-duce journalistic-intensive content. We also build a unique dataset on French newspapers between 1960 and 1974 and perform a difference-in-differences analysis exploiting the introduction of advertising on television, which affected national newspapers more severely than local ones. We find robust evidence of a decrease in the amount of journalistic-intensive content produced and the subscription price.
"Tax Revenues and the Fiscal Cost of Trade Liberalization, 1792-2006" with Lucie Gadenne, Explorations in Economic History, 2018, 70: 1-24.
This article examines the impact of trade liberalization on government revenues. Using a new dataset on tax revenues for 130 countries between 1792 and 2006, we identify ninety-nine episodes of trade liberalization defined as a large fall in trade tax revenues not accompanied by a decrease in trade. Seven took place before World War One, seven in the interwar period, eighteen in the Bretton Woods period and the remainder after 1970. We examine the extent to which countries were able to recover the tax revenues lost by liberalizing trade by using other sources of revenue. We find that historical (pre-1970) trade liberalization episodes were unlikely to be accompanied by decreases in tax revenues, especially during the Bretton Woods era. In the recent period however, over 40% of the developing countries in our sample experience a fall in total tax revenues that lasts more than ten years after an episode of trade liberalization. Overall, trade liberalization led to larger and longer-lived declines in tax revenues in developing countries since 1970 than in today’s rich countries in the 19th and 20th centuries. Results are similar when we consider government expenditures, suggesting decreases in trade tax revenues negatively affect governments’ capacity to provide public services in many developing countries.
"Taxation, Corruption and Growth" with P. Aghion, U. Akcigit and W. Kerr, European Economic Review, 2016, 86: 24-51.
"The Long-Term Effects of the Printing Press in sub-Saharan Africa" with Valeria Rueda, American Economic Journal: Applied Economics, 2016, 8(3): 69-99.
This article investigates the long-term consequences of the printing press in the nineteenth century sub-Saharan Africa on social capital nowadays. Protestant missionaries were the first to import the printing press and to allow the indigenous population to use it. We build a new geocoded dataset locating Protestant missions in 1903. This dataset includes, for each mission station, the geographic location and its characteristics, as well as the printing-, educational-, and health-related investments undertaken by the mission. We show that, within regions close to missions, proximity to a printing press is associated with higher newspaper readership, trust, education, and political participation.
"Improving 'National Brands': Reputation for Quality and Export Promotion Policies" with Dorothée Rouzet, Journal of International Economics, 2015, 95(2): 274-290.
This paper studies the effect of firm and country reputation on exports when buyers cannot observe quality prior to purchase. Firm-level demand is determined by expected quality, which is driven by the dynamics of consumer learning through experience and the country of origin's reputation for quality. We show that asymmetric information can result in multiple steady-state equilibria with endogenous reputation. We identify two types of steady states: a high-quality equilibrium (HQE) and a low-quality equilibrium (LQE). In a LQE, only the lowest-quality and the highest-quality firms are active; a range of relatively high-quality firms are permanently kept out of the market by the informational friction. Countries with bad quality reputation can therefore be locked into exporting low-quality, low-cost goods. Our model delivers novel insights about the dynamic impact of trade policies. First, an export subsidy increases the steady-state average quality of exports and welfare in a LQE,but decreases both quality and welfare in a HQE. Second, there is a tax/subsidy scheme based on the duration of export experience that replicates the perfect information outcome. Third, a minimum quality standard can help an economy initially in a LQE moving to a HQE.
"Improving upon the World Bank's Country Policy and Institutional Assessment: A New Performance Indicator Based on Aid Effectiveness", Journal of Globalization and Development, 2014, 5(2): 213-233.
"Is Charitable Giving Political? New Evidence from Wealth and Income Tax Returns" with Malka Guillot (ETH Zürich).
What are the determinants of philanthropic and political giving? Are donations to charities and parties driven by the same incentives? In this article, we use new administrative household panel data to quantify empirically the motivations for giving, depending on donors’ characteristics and tax incentives. Our dataset includes all the households filling their income tax and/or their wealth tax returns in France between 2006 and 2019. In France, both charitable and political donations benefit from income tax deductions, but only the charitable ones are eligible to the wealth tax credit. We exploit the 2018 wealth tax reform – a change in the taxable base that led to a drop by two third in the number of liable households – to estimate the cross-price elasticity of charitable and political giving at different levels of the income and wealth distributions. We provide new evidence of substituability between charitable and political donations: according to our estimates, a one-percent increase in the price of charitable giving leads to a 19 to 25% increase in political donations. The magnitude of this effect is particularly strong among the the top 20% of the wealth taxpayers.
“Heroes and Villains: The Effects of Combat Heroism on Autocratic Values and Nazi Collaboration in France” with Anna Dagorret (Paris School of Economics), Saumitra Jha (Stanford University), and Pauline Grosjean (UNSW Business School). Submitted.
Can heroes legitimize strongly-proscribed and repugnant political behaviors? We exploit the purposefully arbitrary rotation of French regiments to measure the legitimizing effects of heroic credentials. 53% of French line regiments happened to rotate under a specific general, Philippe Pétain, during the pivotal WWI battle of Verdun (1916). Using recently- declassified intelligence data on 95,314 individuals, we find the home municipalities of regiments serving under Pétain at Verdun raised 7% more Nazi collaborators during the Pétain- led Vichy regime (1940-44). The effects are similar across joining Fascist parties, German forces, paramilitaries that hunted Jews and the Resistance, and collaborating economically. These municipalities also increasingly vote for right-wing parties between the wars. The voting effects persist after WWII, becoming particularly salient during social crises. We argue these results reflect the complementary role of the heroes of Verdun in legitimizing and di↵using the authoritarian values of their former leader.
This paper received the Oliver Williamson Best Conference Paper Award, for the best paper delivered at the SIOE (Society for Institutional & Organizational Economics) 2020 Conference.
CEPR Discussion Paper #15613.
"The Heterogeneous Price of a Vote: Evidence from Multiparty Systems, 1993-2017" with Yasmine Bekkouche (Université Libre de Bruxelles) and Edgard Dewitte (Sciences Po Paris). R&R Journal of Public Economics.
What is the impact of campaign spending on votes? Does it vary across election types, political parties or electoral settings? Estimating these effects requires comprehensive data on spending across candidates, parties and elections, as well as identification strategies that handle the endogenous and strategic nature of campaign spending in multiparty systems. This paper provides novel contributions in both of these areas. We build a new comprehensive dataset of all French legislative and UK general elections over the 1993-2017 period. We propose new empirical specifications, including a new instrument that relies on the fact that candidates are differentially affected by regulation on the source of funding on which they depend the most. We find that an increase in spending per voter consistently improves candidates' vote share, both at British and French elections, and that the effect is heterogeneous depending on candidates' party. In particular, we show that spending by radical and extreme parties has much lower returns than spending by mainstream parties, and that this can be partly explained by the social stigma attached to extreme voting. Our findings help reconcile the conflicting results of the existing literature, and improve our understanding of why campaigns matter.
Technological innovations in content delivery, such as the advent of broadcast television or of the Internet, threaten local newspapers’ ability to bundle their original local content with third-party content such as wire national news. We examine how the entry of television – with its initial focus on national news – affected local newspapers as well as consumer news diets in the United States. We develop a model of local media and show that entry of national television news could reduce the provision of local news. We construct a novel dataset of U.S. newspapers’ economic performance and content choices from 1944 to 1964 and exploit quasi-random variation in the rollout of television to show that this new technology was a negative shock in both the readership and advertising markets for newspapers. Newspapers responded by providing less content, particularly local news. We tie this change towards increasingly nationalized news diets to a decrease in split-ticket voting across Congressional and Presidential elections.
"Social Media and Newsroom Production Decisions" with Nicolas Hervé (INA) and Béatrice Mazoyer (INA)
Social media affects not only the way we consume news, but also the way news is produced, including by traditional media outlets. In this paper, we study the propagation of information from social media to mainstream media, and investigate whether news editors are influenced in their editorial decisions by stories popularity on social media. To do so, we build a novel dataset including a representative sample of all tweets produced in French between July 2018 and July 2019 (1.8 billion tweets, around 70% of all tweets in French during the period) and the content published online by about 200 mainstream media during the same time period, and develop novel algorithms to identify and link events on social and mainstream media. To isolate the causal impact of popularity, we rely on the structure of the Twitter network and propose a new instrument based on the interaction between measures of user centrality and news pressure at the time of the event. We show that story popularity has a positive effect on media coverage, and that this effect varies depending on media outlets’ characteristics. These findings shed a new light on our understanding of how editors decide on the coverage for stories, and question the welfare effects of social media.
"It Takes Money to Make MPs : New Evidence from 150 Years of British Campaign Spending" with Edgard Dewitte (Sciences Po Paris)
We study electoral campaigns over the long run, through the lens of their spending. In particular, we ask whether changing media technologies and electoral environments impacted patterns of spending and their correlation with electoral results. To do so, we build a novel exhaustive dataset on general elections in the United Kingdom from 1857 to 2017, which includes information on campaign spending (itemized by expense categories), electoral outcomes and socio-demographic characteristics for 69, 042 election-constituency-candidates. We start by providing new insights on the history of British political campaigns, in particular the growing importance of advertising material, including via digital means, to the detriment of paid staff and electoral meetings. We then show that there is a strong positive correlation between expenditures and votes, and that overall the magnitude of this relationship has strongly increased since the 1880s, with a peak in the last quarter of the 20th century. We link these transformations to changes in the conduct of campaigns, and to the introduction of new information technologies. We show in particular that the expansion of local radio and broadband Internet increased the sensitivity of the electoral results to differences in campaign spending. These results encourage greater contextualization in the drafting of campaign finance regulations.
"Hosting Media Bias: Evidence from the Universe of French Television and Radio Shows, 2002-2020" with Moritz Hengel (Sciences Po Paris), Nicolas Hervé (INA) and Camille Urvoy (Sciences Po Paris). Draft available upon request.
Editorial independence is a central normative principle in media policy and journalism. Yet, in practice, empirical evidence shows that an important number of media outlets tend to slant information, and that this bias is partly driven by owners. Using new data on hosts and guests in all French radio and television shows between 2002 and 2020, we study the role played by journalists in shaping media slant, and how they adapt to changes in ownership. We measure media bias using the speaking time share of politicians from different parties, as well as the one of politically-involved non-politician guests. Using France 2 – the leading public television channel – as a benchmark, we document that most channels systematically grant more speaking time to right- (left-)wing guests, regardless of the ruling party. We also find substantial variation in slant across ownership groups. We next use hosts working on several channels or stations that belong to different owners and show that, controlling for demand- and supply-driven bias, journalists themselves slant media content. We finally study how changes in ownership affect media bias in a difference- in-differences framework using several takeovers that occurred during our time period. We find that ownership impacts media slant: in particular, after Vincent Bolloré’s – one of the top French media tycoons – takeover of CNews, Canal+ and C8, these television channels started to grant more speaking time to far-right politicians. Concurrently, we show that ownership changes trigger a decline in the broadcast time share of hosts who were working for the channel before the takeover.
“The Rising Demand for Representation: Lessons from 100 Years of Political Selection in the UK” with Edgard Dewitte (Sciences Po Paris).
Do citizens care about descriptive representation? The rise of populisms and their rejection of traditional political elites point towards changes in citizens’ preferences regarding the characteristics of their politicians. In this article, we build a novel dataset on the profiles of all the candidates at the United Kingdom General Elections since 1918, and of the constituencies in which they run, in order to analyze the relationship between political selection and electoral outcomes. Our dataset covers 30,590 unique candidates running in 13,022 election-constituencies, for which we have information on gender, age, education as well as previous occupations. Building a "representation gap" index, we start by documenting the changes in patterns of representation and their impact on electoral participation. At the constituency level, we show that, since the early 2000s, turnout is lower the more candidates differ from their electorate in terms of education and occupation status - while the opposite was true over the XXth century. Similarly, we find that representativeness is now associated with an electoral advantage at the candidate level. These findings lead to a reassessment of the quality-representation trade-off; and of the reasons behind the persistence in the gap between political elites and those they represent.
Do campaign contributions influence politicians? In this article, we study the impact of corporate donations on ideology and political discourse. We construct a novel dataset that combines the campaign manifestos issued by every candidate running for a seat in the French parliament with data on the amount and origin of their campaign contributions. We exploit an exogenous historical shock on corporate donations to estimate their causal impact on the content of campaign communication. Combining a difference-in-differences approach with computational text analysis, we show that receiving more donations from small and local corporate donors encourages candidates to advertise their local presence over national politics during the campaign. We also find evidence that donations lead candidates from extreme parties to moderate their rhetoric—including shifts in the policy topics they advertise. Our findings may reflect a "quid-pro-quo effect" between donors and politicians, but they may also result from an "electoral effect": receiving expressive contributions from corporate donors affects the content of campaign messages because it changes candidates' perception of which issues matter most to voters. According to our findings, campaign finance regulations may alter the information made available to voters through their impact on candidates' rhetoric.
"Electoral Importance and Media Consumption: Quasi-experimental Evidence and New Data from India" with Guilhem Cassan (University of Namur) and Francesca Jensenius (Norwegian Institute of International Affairs). Draft available upon request.
What are the determinants of news media consumption? In this paper, we investigate whether it is determined by political motives. We build a new panel dataset on Indian publications at the city level between 2002 and 2017. We exploit the 2008 delimitation of the Assembly Constituencies – an exogenous change in the electoral importance of cities across India – to causally identify the relationship between relative electoral importance and news media consumption. Using a difference-in-differences approach, we compare change in the supply and demand of news of cities whose electoral importance increased compared to cities whose electoral importance did not. We show that media markets whose electoral importance increases see an increase in their total newspaper circulation per capita. We discuss how this political motive can be decomposed into media supply and media demand.
" Democratization, Media Supply and Media Bias: Evidence from the USA", with Guilhem Cassan (University of Namur), Eric Melander (University of Namur), and James Snyder (Harvard University).
“Payroll and Inequality within the Newsroom: Evidence from France, 1936-2016”
“Estimating the Production and Demand for Online News: Micro-Level Evidence from the Universe of French News Media” with Nicolas Hervé (INA) and Marie-Luce Viaud (INA)
“Word Embeddings for Topic Detection and Tracking” with Béatrice Mazoyer (INA), Nicolas Hervé (INA), Marc Evrard (INA) and Céline Hudelot (Ecole Centrale Paris).
"A French Corpus for Event Detection on Twitter" with B. Mazoyer, N. Hervé and C. Hudelot, Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020).
“Représentations lexicales pour la détection non supervisée d’événements dans un flux de tweets : étude sur des corpus français et anglais” with B. Mazoyer, N. Hervé and C. Hudelot, Extraction et Gestion des Connaissances (EGC 2020).
Dans cet article, nous nous intéressons aux approches récentes de plongements lexicaux en vue de les appliquer à la détection automatique d’événements dans un flux de tweets. Nous modélisons cette tâche comme un problème de clustering dynamique. Nos expériences sont menées sur un corpus de tweets en anglais accessible publiquement ainsi que sur un jeu de données similaire en français annoté par notre équipe. Nous montrons que les techniques récentes fondées sur des réseaux de neurones profonds (ELMo, Universal Sentence Encoder, BERT, SBERT), bien que prometteuses sur de nombreuses applications, sont peu adaptées pour cette tâche, même sur le corpus en anglais. Nous expérimentons également différents types de fine-tuning afin d’améliorer les résultats de ces modèles sur les données en français. Nous proposons en- fin une analyse fine des résultats obtenus montrant la supériorité des approches traditionnelles de type tf-idf pour ce type de tâche et de corpus.
“Real-Time Collection of Reliable and Representative Tweets Datasets Related to News Events” with B. Mazoyer, C. Hudelot and M.-L. Viaud, Proceedings of the First International Workshop on Analysis of Broad Dynamic Topics over Social Media (BroDyn 2018) co-located with the 40th European Conference on Information Retrieval (ECIR 2018), 2018, 23–34.
"Asymmetric Information, Rent Extraction and Aid Efficiency", 2009.
Official Development Aid flows are volatile, non-predictable and not delivered in a transparent way. All these features reinforce asymmetric information between the citizens and the recipient government about the amount of aid flows received by developing countries. This article uses a political economy model of rent extraction to show how this asymmetry (i) encourages rent extraction by kleptocratic regimes, thus reducing aid efficiency, and (ii) increases the negative impact of aid volatility. It identifies a new channel - the "asymmetric information" channel - through which aid volatility is costly for recipient countries. The empirical relevance of the model is confirmed on a panel data of developing countries. Using various specifications and econometric methods, and developing new yearly estimates of aid volatility, I show that (i) introducing more information increases aid efficiency, that (ii) the negative impact of aid volatility on aid efficiency vanishes once one controls for information, and that (iii) this positive impact of information does not come from the fact that more transparent countries tend to have better institutions.
"Aid Volatility and Macro Risks in Low-Income Countries" with Eduardo Borensztein, Daniel Cohen and Cécile Valadier, OECD Development Centre 273, 2008.