POLITICAL POWER OF THE PRESS IN THE WEIMAR REPUBLIC∗

BANG DINH NGUYEN†

Abstract. This paper studies the political power of the press in the Weimar Republic (1918–1933). Using a dataset of 135 daily newspapers in Berlin during this period, I ex- amine how changes in the circulation figures of partisan newspapers affected voter turnout, party’s vote shares and mass political polarization at three levels: federal, state and city council elections. My empirical strategy exploits the variation in the development of the urban commuter rail system across 20 historical districts of Berlin, which influenced the circulation of newspapers in each district. My analysis shows that an increase in newspaper circulation significantly induced higher turnout in 14 outer districts of Berlin, with ten thou- sand additional copies of newspapers increasing turnout by approximately 0.11 percentage points, and gave rise to a higher degree of mass polarization. In addition, ten thousand additional copies of newspapers supporting a party led to 0.31 percentage point increase in its vote share in federal elections. The effect was stronger for federal elections than for state elections and city council elections. Most importantly, the electoral influence of news- papers was driven by tabloids and mass newspapers, and not by elite political newspapers. Among the anti-Republican press, the right-wing press exerted a strong influence in federal elections, whereas the power of the communist press was negligible.

Keywords: Press, Newspapers, Politics, Election Outcome, Railways JEL classification: D72, L82, N44, N74, N84

Date: November 27, 2019. Click here for the most recent version. †Stern School of Business, New York University. Email: [email protected]. ∗I thank my advisors – Petra Moser, Walker Hanlon and Lu´ısCabral – for their guidance and encouragement. I also received helpful comments from Felipe Alves, Roxana Mihet, Juergen Reichert, Lawrence White and participants at the NYU Applied Micro Lunch Workshop, EHES Congress 2019 and EHA Meeting 2019.

1 1. Introduction

The question of how mass media affects political outcomes has attracted a great deal of scholarly attention recently. A large strand of literature finds significant persuasive effects of modern media outlets on party’s vote shares (DellaVigna and Kaplan, 2007; Gerber et al., 2009; Enikolopov et al., 2011).1 Several scholars attempt to study the political influence of newspapers in the 19th and early 20th centuries when print media was still the most predominant form of media. Gentzkow et al.(2011), for example, conduct a historical study of the effect of US daily newspapers on electoral politics from 1869 to 1928 and find that the existence of partisan newspapers increased voter turnout but had no significant effect on party’s vote shares. How robust are these results to different political systems and different institutions in other countries? This paper examines the case of the Weimar Republic, the fourteen-year period starting in 1918 and ending in 1933, when Hitler rose to power. In Weimar Germany, the national legislative body – the Reichstag – was filled with a large number of middle-sized and small political parties. A pure proportional representation system was implemented, specifying that each party received the same percentage of seats in the Reichstag as the percentage of its vote share. This stood in stark contrast to the United States, where a winner-take-all system has been used for a long time as the basis for political decision-making in the two-party system. In addition, basic attitudes towards journalism in the Weimar Republic were strikingly different from those in Anglo-Saxon countries (Eksteins, 1975). Whereas in the United States and Britain the most important goal for an aspiring journalist was to report facts with an impersonal, objective approach, in Germany journalists were encouraged to express opinions and judgments in their articles, to become participants in controversies rather than to remain spectators. Arthur Koestler – foreign editor of the B.Z. am Mittag – reflected in his memoirs: [German journalism’s] starting point was the correspondent’s Weltanschauung, and the political philosophy of the paper for which he worked. His job was not to report the news and facts. . . but to use facts as pretexts for venting his opinions and passing oracular judgement. ‘Facts,’ a famous German editor said, ‘are not fit for the reader when served raw; they had to be cooked, chewed and presented in the correspondent’s saliva.’ (Koestler, 1952) Another distinct characteristic of the Weimar press was the fragmentation and decentraliza- tion in the newspaper market. Most German newspapers were private enterprises owned by individual families. Each newspaper focused on its own city, and there were neither national newspapers as in Britain (e.g., The Times, The Guardian, The Daily Mail), nor chains owned

1In contrast to other results, Falck et al.(2014) find that the availability of the Internet reduced voter turnout and did not influence party’s vote shares in Germany for the election cycle 2004–2008.

2 by one company as in the United States (e.g., the Hearst company had a nationwide string of 28 newspapers by the mid 1920s). Taking the political system, the attitudes towards journalism and the newspaper market structure into consideration, this paper studies the interplay between the press and electoral outcomes in Berlin during the Weimar Republic. This period in German history provides an excellent context to study the political role of the press, since in the 1920s newspapers were the predominant medium of mass communication in Germany and constituted the only available window on politics. Furthermore, German newspapers had a long tradition of adhering to a political cause. The Revolution of 1848 resulted in the foundation of a large number of newspapers with a clearly defined political agenda. After World War I, most German newspapers had to reconsider their politics. Each newspaper listed its political stance and explicitly encouraged readers to vote for a particular party on election days. See Figure1 for an example. Surprisingly, studies of the Berlin press have not been able to establish a connection between newspapers and electoral behavior (Fulda, 2009). This result is worth reconsidering, since Weimar contemporaries claimed that they derived their political knowledge through information mediated by the press, and reading newspapers was their everyday activity on public transport. As Georg Bernhard – editor-in-chief of the Vossische Zeitung – put it, a German viewed his favorite newspaper as “not merely a source of information, but also an organ of instruction. . . The German believes what his paper tells him. Men to whom much space is devoted are to him great men.” (Bernhard, 1929) A careful examination of these studies reveals that their primary focus was solely on elite political newspapers. These publications were often regarded as official party organs, primarily concerned with high politics and characterized by high subscription fees and low circulation. A remarkable feature of the Berlin press in this period is, however, the rising popularity of two alternatives to these old-fashioned political papers: mass newspapers and tabloids. By investigating a comprehensive dataset of the Berlin daily press which consists of all these three types of newspapers, I re-examine the relationship between the press and elections in the Weimar Republic. This quantitative approach complements previous historical studies, whose the results had been based on the analysis of case studies, commentaries, letters and papers from individual politicians or editors. To estimate the causal effect of newspaper circulation on political outcomes, any potential endogeneity problem needs to be addressed. For example, a district with a larger volume of newspapers would have a higher turnout due to a higher level of education whose data may not be available, instead of a direct causal effect of newspapers. As an empirical strategy, I exploit the variation in the development of the urban rail system across 20 historical districts

3 of Berlin. In particular, I use the number of railway stations in a district as an instrumental variable for the circulation of newspapers in that district. The growth of Berlin railway system played an essential role in improving the collection of news, expediting distribution and as well as lowering the delivery cost of newspapers; therefore, it can serve as a good proxy for newspaper circulation. Preston(1999) explains that railways allowed a quicker and more efficient distribution network to be created, especially in rural areas, as newspapers could quickly be sent in bulk to agents in towns with a railway station, and then be sent, probably by road, to agents in nearby settlements. See Figure2 for different channels of distributing newspapers from a publishing house to the readers, where railroads were a crucial means of transportation. Jones(1993, pp.105-106) studies the history of journalism in Wales and asserts that the development of the railway network was the “single most important contribution to the growth of a mass newspaper market”. This also reflected the situation in Weimar Berlin. New railway stations not only improved the distribution of newspapers but also fostered new venues for selling them. Bookshops at railway stations are a case in point. They had became integral features of the modern landscape from the beginning of the 20th century. The most popular reading materials to be sold in railways stations were newspapers and journals, rather than books, which are generally heavy and take up a great deal of space. Remarkably, these bookshops strived to cater to readers from all social strata.2 These distribution outlets, which saw themselves as politically neutral entities addressing an anonymous target readership without any regard for religion, status or political affiliation, and which set out to satisfy all the reading requirements of ‘Mr Everyman’ (Herr Jedermann), sought to appeal to the broadest common denominator of the varied public passing through railway stations (Reuveni, 2002). Already before the First World War, street vendors had tried to sell newspapers directly in public transport (Fischer and F¨ussel, 2007, p.541). In 1907, the company Straßenbahn- Zeitungsvertrieb obtained the permission to put uniformed newsboys in the Berlin railways for some fees (Pachtgeb¨uren). They got in at one station and got off at the next station, selling daily newspapers, weekly publications and comic strips (Haug, 2007, pp.243-244). See Figure3 for an example. As the railway network expanded, street sales of newspapers also became more widespread, newspapers were able to reach more readers in areas where a station had not previously existed. In addition to the supply factors mentioned above, the relationship between newspaper circulation and the development of Berlin railway system can also be viewed from a demand

2The distribution of socialist newspapers in railway stations was forbidden until November 1918 (Reuveni, 2006b, p.107). The neutral stance during the Weimar era was also encouraged by the leases between the Railway Administration and the railway bookshops.

4 perspective. The railway network not only facilitated the transfer of information and goods from one neighborhood to another but also led to the emergence of a new readership – the traveling public. A longer train journey created a demand among Berlin commuters for cheap reading materials, and newspapers, which became more and more widespread, were an easy choice. Moreover, it can be argued that the variation in the number of stations in a district were not correlated with the election outcomes in that district. The plans to expand and improve the commuter rail network in Berlin had already been determined in the Imperial era, but most of which were postponed until the Weimar period due to the war. For example, the construction of the North-South railway (Nord-S¨ud-Bahn) connecting Wedding to Tempelhof and Neuk¨ollnhad started in December 1912, was delayed during the war and only resumed in 1919. To study electoral influence of the newspapers, I examine how the variation in newspaper circulation affected electoral outcomes, in particular, voter turnout, party’s vote share and mass political polarization. My analysis shows that an increase in newspaper circulation significantly increased voter turnout in the outer districts of Berlin – with ten thousand additional copies of newspapers increasing turnout by 0.11 percentage points – and gave rise to a higher degree of electorate polarization. Furthermore, an increase of ten thousand copies in the circulation of newspapers that favored a political party led to, on average, 0.31 percentage point increase in that party’s vote share in federal elections. However, partisan newspapers benefited parties more in federal elections than in state elections or in city elections. In addition, I compare the effects across different types of newspapers and find that mass papers and tabloids had a stronger influence than did elite political papers. These results confirm the results found in historical case studies of Berlin newspapers which centered on political publications, and help identify the channel through which newspapers affected election outcomes. Finally, I study how the strength of the press influence varied across different political tendencies of newspapers. Among the anti-Republican press, the right-wing press exerted a large influence on electoral outcomes, whereas the influence of the communist press was negligible. In what follows I discuss the related literature and provide a brief overview of the political and historical context that is relevant for my analysis. I then elaborate on my data and my analysis. Some possible mechanisms for my findings are discussed, followed by concluding remarks.

5 2. Related Literature

The market for news has attracted research in economics literature only very recently. A large part of this literature has focused on the effects of newspapers on political outcomes.3 Snyder and Str¨omberg(2010) estimate the impact of press coverage on citizen knowledge, politicians’ actions and policy for the period 1991–2002. They find that in areas with a higher degree of congruence between the news market and congressional districts, voters had more information and were more likely to turn out to vote. Using the data on entry and exit of US daily newspapers from 1869 to 1928, Gentzkow et al.(2011) estimate the effect on political participation, party votes share and electoral competitiveness. By comparing the election outcome of a county that experienced a change in the number of newspapers to the county that did not in the same state and year, the authors find that newspapers had a robust positive effect on turnout but not on vote shares, and did not systematically benefit or hurt the incumbents. This study, however, centers on newspapers in the United States where several political and social aspects were different from the Weimar Republic. In addition, the focus of their analysis was on presidential elections. Following a similar approach, other studies have explored the political influence of newspapers in other institutional settings outside the United States. Drago et al.(2014), for example, show that newspaper entry increased turnout in Italian municipal elections but did not affect the selection of politicians. Cag´e(2017) studies the impact of increased media competition on the quantity and quality of news and on political participation in France. Contrary to the results in the two papers above, the author shows that entry reduced voter turnout at local elections and also sharply lowered circulation. In addition to newspapers, a large strand of literature finds significant persuasive effects of other types of partisan media. Str¨omberg(2004) studies the effect of radio on public spending and shows that the US counties with more radio listeners were associated with a higher level of electoral participation and also received more relief funds during the New Deal. Gentzkow(2006) uses variation in the timing of television’s introduction across US counties and finds a negative effect on turnout. Among the studies that examine the political effect of media in the Weimar Republic, Adena et al.(2015) is the most closely related to my paper. The authors exploit the changes in the content of radio broadcast and the variation in radio exposure to study how radio affected political support for the Nazi party. In their paper, the radio served as a channel through which political information reached the voters, whereas in this paper, the railway stations played an essential role in exposing citizens to newspapers and consequently to the political messages contained in these newspapers. The authors find that the organization which controls media content can play a significant role, either in safeguarding an unconsolidated democracy or in contributing to its collapse. However,

3See Prat and Str¨omberg(2013) and Sobbrio(2014) for detailed surveys.

6 their paper focuses only on the Nazi party and on the later years of the Weimar Republic, i.e., only after the radio had really taken off. In my paper, I look at eight dominant parties throughout the 14 years of the Weimar era. Scholars have attempted to explain media bias from both demand and supply perspec- tives. There is evidence on demand-driven media bias, i.e., media content being tailored to the beliefs of consumers, which has been studied both theoretically (see Mullainathan and Schleifer, 2005; Gentzkow and Shapiro, 2006) and empirically (see Hamilton, 2006; Gentzkow and Shapiro, 2010). Puglisi(2011) studies the news from The New York Times between 1946 and 1997 and finds that the paper had a systematic issue selection bias favoring the Demo- cratic candidate during presidential campaign. Gentzkow et al.(2014) show that readers preferred news that confirms their prior beliefs and newspapers used their political orienta- tion to differentiate from competitors. With regard to how readers would be influenced by media bias, Chiang and Knight(2011) show that a newspaper’s party affiliation need not affect expected vote shares, since rational voters will take an outlet’s bias into account in updating their beliefs. Similarly, Gentzkow et al.(2011) argue that explicit party affiliations in newspapers made it easier for the readers to filter bias. My paper, however, belongs to the literature on supply-driven media bias, i.e., the existence of partisan newspapers was the result of the preferences of the journalists and the political ideologies of the publishers (Groseclose and Milyo, 2005; Gentzkow and Shapiro, 2010; Durante and Knight, 2012) or of the incentives to differentiate to attract readers with extreme ideologies (Chiang, 2010). In addition to voter turnout and party’s vote share, I examine the effect of newspapers on mass political polarization. In his seminal work, Sartori(1976) defines political polarization broadly as the ideological distance between candidates, parties or voters. Indeed, the notion of “polarization” is not consistently defined in the political science literature. In this paper, I focus my attention on mass polarization, also known as popular polarization, which refers to the polarization in the electorate. For a multiple-party system, Pelizzo and Babones(2003) suggest a measure for polarization by summing up the vote shares of all of the extreme, bipolar, anti-system parties then subtracting the vote share of the incumbent. This index captures party system dynamics, i.e., the extent to which the vote is dispersed along the left- right spectrum, as well as two important features of polarization: a decrease in the electoral strength of the incumbent and an increase in the electoral strength of the anti-system parties. Applying that index to four polarized pluralist party systems, Pelizzo and Babones(2007) find that the worsening economic conditions led to a higher level of polarization, including the case of the Weimar Republic. With regard to the United States, there is mixed evidence that partisan media polarize Americans. Prior(2013) identifies some certain problems in linking the growth of partisan media and the increase in mass polarization and argues that the growth in mass polarization through increasingly partisan media can only be applied to a

7 minority of Americans. Darr et al.(2018) study the closure of 110 newspapers in the United States from 2009 to 2012 and show that after local newspapers closed, political polarization among voters, measured by split-ticket voting, increased.4 This is different from the result in my analysis which shows that a larger volume of newspapers contributed to a higher degree of mass polarization.

3. Historical Context

In this section, I review and discuss the important historical features of the Weimar Re- public that are relevant to the following analysis.

3.1. Political Landscape. The Weimar Republic was proclaimed on November 9, 1918 by the Social Democrat Philipp Scheidemann. The first full-fledged cabinet drew from the three leading parties – the Social Democrats (SPD), the Democrats (DDP) and the Center (Zen- trum), which together formed the Weimar Coalition. Under the 1919 Weimar Constitution, Germany became a democratic republic, which was designed to allow a multiplicity of voices and a diversity of points of view. Elections to the Reichstag were by universal suffrage starting at age 20, and for the first time women were granted the right to vote and to stand as candidates. A pure form of proportional representation replaced the previous first-past-the-post election system in the Kaiserreich – each party was awarded one seat in the Reichstag for every 60,000 votes nationwide. The number of seats in the Reichstag from 1919 to 1933 varied between 423 and 647. In 1919 election to the National Assembly (Nationalversammlung) took place. Subse- quently there were eight elections to the Reichstag: June 1920, May 1924, November 1924, May 1928, September 1930, July 1932, November 1932 and March 1933.5 The President was elected for a seven-year term by popular vote using a two-ballot system. Three presidential elections took place in this period: in 1919, 1925 and 1932. A remarkable feature in the Reichstag was its plethora of divergent political parties. It was quite common for ballot papers to contain more than 30 parties and candidates. See Figure4 for a ballot paper used in the election to the Reichstag in 1932. A voter would not vote for a specific candidate, but vote for a party. Candidates were presented on a party list, with the order of candidates determined by the party leadership in each district. As a consequence of the proportional representation system, the Reichstag consisted of more than a dozen different parties from across the political spectrum. For example, in the 1920 election, 14 different parties won Reichstag seats and in 1930, the number of winning

4Split ticket voters are divided in their party allegiances, for example, they vote for a Democratic presidential candidate and a Republican senatorial candidate. 5A detailed description of these elections can be found in Elections in the Weimar Republic from Deutscher Bundestag.

8 parties was 17 (Vogel et al., 1971). The number of parties which had at least one seat in the parliament in between 1919 and 1933 was 29. Proportional voting and the abundance of parties made it impossible for one or even two parties to dominate the Reichstag. In this paper, I examine eight biggest parties that existed during this period: from the left to the right of the political spectrum, the Communist Party (KPD), the Independent Social Democratic Party (USPD)6, the Social Democratic Party (SPD), the Democratic Party (DDP), the Center Party (Z ), the People’s Party (DVP), the National People’s Party (DNVP) and the National Socialist Workers’ Party (NSDAP).7 In addition to federal elections, I examine Prussian state elections (Landtag) and Berlin city council elections (Stadtverordnetenversammlung). State elections took place according to state electoral laws, which varied from state to state. In Prussia, as well as in Baden, Mecklenburg-Schwerin, Oldenburg and Thuringia, state elections were based on the Reich election law (Leicht, 2005). There were six elections to the Prussian Landtag and six city council elections in Berlin during the Weimar era. From 1919 to 1933, the number of seats in the Prussian Landtag varied between 402 and 476, and the number of Berlin city councillors ranged between 93 and 225.

3.2. Berlin. Berlin was the capital of the German Reich, the capital of Prussia – the Repub- lic’s largest state – and also Germany’s most important industrial city. On April 27, 1920, the Prussian government passed the Greater Berlin Act (Groß-Berlin-Gesetz), expanding the size of Berlin from 66 to 883 square kilometers. Greater Berlin now consisted of eight towns, 59 rural communities (Landgemeiden) and 27 estate districts (Gutsbezirke). It was then sub-divided into 20 districts (Verwaltungsbezirke) (See Figure5). My analysis is at the district level, and I focus my attention on elections after April 1920, when these 20 districts had been established. I refer to the six original districts of old Berlin (the darkly shaded dis- tricts in Figure5) as inner districts. These include Mitte, Tiergarten, Wedding, Prenzlauer Berg 8, Friedrichshain and Kreuzberg 9. We refer to the new 14 districts as outer districts. With a population of 4,024,165 in 1925, Berlin was the third biggest city in the world after New York and London. More importantly, during the Weimar Republic, Berlin featured more newspapers than any other city. Each Berlin hour throws millions of newspaper pages onto the streets, into houses, into the administration, into the directors’ offices of banks, into branch offices, into factory offices, into taverns and into the theatre. They flood public transport. . . they drown the parks and they are being transported by

6The USPD was dissolved in 1931. 7A detailed description of these eight parties can be found in The Political Parties in the Weimar Republic from Deutscher Bundestag. 8Originally named Prenzlauer Tor, changed to Prenzlauer Berg in 1921. 9Originally named Hallesches Tor, renamed to Kreuzberg on September 27, 1921.

9 newspaper planes beyond mountains, forests and seas: politics, economics, traffic, technology, stock market, sport, art fill and shake the air. . . each hour with loud and novel news, which the press is giving speedy wings (Fulda, 2009).

In 1924, the Federal Foreign Office (Ausw¨artigeAmt) drew up a list of the most important German dailies and their party affiliations. The list consisted of 66 newspapers, more than a third of which were Berlin papers (Fulda, 2009). Furthermore, Berlin hosted the three biggest publishing houses in the country – Mosse (liberal left), Ullstein (liberal right) and Scherl (the extreme or national right) (Lerg, 1989). In the early 1930s, a large share of reading materials originated from these publishing houses. The press quarter was located in the heart of the city, between Potsdamer Platz and Anhalter Bahnhof railway station, and separated from each other by only a few blocks. The main publishers were based in Zimmerstraße, Kochstraße, Hedemannstraße and Jerusalemstraße, and it was in these streets that the barricade fights of January revolt of 1919 took place. Fritzsche(1996) asserts, “no city could have supported so many newspapers had city people not browsed their way through the diverse selections of the press. If Berliners subscribed to a morning paper, in most cases the Morgenpost, they also purchased paper at street-corner kiosks and perused any one of the latest editions lunching in caf´es,waiting for buses, and riding streetcars.”

3.3. The Market for Newspapers. A great number of research papers on mass culture in the Weimar Republic center on radio and film. However, throughout the 1920s newspapers dominated other forms of mass communication in Germany. Radio broadcasting started to become popular only at the end of the 1920s, with 3.5 million registered listeners by 1930. Television was introduced to the public for the first time at the fifth Große Deutsche Funkausstellung (a radio exhibition) in 1928. With regard to films, 350 million cinema tickets were sold in 1929. In the same year, daily circulation of newspapers were about 20 million copies. Moreover, newspaper reading was the most popular leisure activity and an important cornerstone of mass entertainment in this period; also each newspaper tended to be read by more than one person. As Fulda(2009) puts it, “if the 1930s were to become the decade of film and radio in Germany – and even this is open to doubt – the 1920s were undoubtedly the decade of the press.” In the nineteenth century, German newspapers had been small, distinctively political en- terprises with a limited public, catering mainly to elites and focusing heavily on national politics. In 1874, the government passed a law to abolish the prohibitive tax (Stempels- teuer) on the press. In 1918, the abolition of censorship in Germany initiated a veritable renaissance of the press. Article 118 of Weimar Constitution proclaimed the freedom of the press: “every German has the right, within the bounds of the general laws, to express his

10 opinion freely in word, print, picture, or in any other manner. . . There is no censorship.”10 Together with innovations in printing technology, the press experienced remarkable growth. In 1928 there were 3,356 newspapers in the German Reich. Compared to today, in 2018, only 681 different newspapers can be found in Germany according to the Federal Association of German Newspaper Publishers (BDZV ). The growth was also fueled by a new concept of commercial newspaper (Generalanzeiger) which emerged in the 1880s. For these papers, the main source of funding came from advertising income rather than from sale income as was the case with the traditional papers. The dependence of advertisements limited a paper’s geographical distribution. In 1927, 56% of all newspapers (1,811 out of 3,241) appeared in municipalities with less than 10,000 residents (Dussel, 2012). The popularity of journals and newspapers in particular, which reached new circulation records during the Weimar period, represented for many German scholars and academic at the time the transition of German society from a book-reading bourgeois society to a newspaper-reading mass society (Reuveni, 2006a, p.205). The German newspaper market in the Weimar Republic was fragmented and decentralized. In this period, the press was not a homogeneous collective but a colorful assortment of very different publishing enterprises. Most German newspapers were private enterprises owned by individual families. In 1926, more than 80 percent were still family-owned (Eksteins, 1975). Many newspapers issued a morning and evening edition from Tuesday to Saturday. In addition, there were Monday, Sunday and midday newspapers. The effects of political polarization on the newspapers were fundamental. Contemporaries could hardly find similar information about a political topic from two different papers. News was reported highly selectively; stories were given a strong slant and were edited according to the papers’ different political outlooks.11 Mosse’s Berliner Tageblatt, Ullstein’s Berliner

10However, the freedom of expression was still subjected to the penal laws, police ordinances, and to the regulations on trade and business practices (Eksteins, 1975, p.70). For example, to prevent extremists from destroying the political order of the Weimar Republic, the Republikschutzgesetze (laws for the protection of the republic) were passed in 1922 and 1930, which gave the authorities the right to suppress newspapers if they expressed contempt for the existing state structure, the national flag, or members of the government, or if they advocated violence (Wilke, 2013). 11Kurt Robitschek – director of Kabarett der Komiker, the most successful and the largest cabaret of the Weimar period – described how Berlin’s politicized newspapers would report a minor incident: a collision between a bicyclist and a dog. “The optimistically liberal Berliner Tageblatt claimed that dog and bicyclist race along the Kurf¨urstendamm,they hurry – despite a little scratch here and there – towards the brilliant future of the German Republic. The pessimistically liberal Vossische Zeitung complained that the appearance of red blood on a black dog with white spots turned the incident into an expression of reactionary politics; it called for more laws to defend the Republic. The nationalist Lokal-Anzeiger claimed that a foreign bicyclist ran over the dog of a retired general. Fifteen years ago the German people would have stood up as one body and would have swept the bicyclist away with ringing manly fury, but today our faithful dogs lie limply on the ground, shattered by the Treaty of Versailles. The Communist Rote Fahn reported: On Kurf¨urstendamm, the pompous boulevard of satiated capitalism on which the proletarian revolution will march against the imperialists in the very near future, a dog attacks a single bicyclist, then all dogs unite against the Soviet Union! Finally the Nazi V¨olkischerBeobachter asserted: Once more one of our party comrades had been

11 Morgenpost, and Scherl’s Berliner Lokal-Anzeiger had the highest circulation of all German newspapers, and their sales centered in Berlin. Due to their dependence on local adver- tisements and their strong emphasis on local news coverage, Berlin newspapers hardly ever found their way west of the Elbe river.12 Fulda(2009, p.9) argues that “the fragmenta- tion of the press into competing and often mutually hostile communication networks was a key feature of Weimar Germany’s political culture.” Another important feature of the Berlin newspaper market in this period is the frequent mergers and disappearance of many partisan newspapers. In general, daily newspapers in Berlin can be divided into three main categories: elite political papers, mass papers and tabloids (see Figure6).

3.3.1. Elite political papers. Elite political papers – also regarded as official party organs – were the traditional old-fashioned papers. These publications were read by German elites (bureaucrats, politicians, military leaders, industrialists, journalists and academics) and arguably had little influence on the voting population. With high subscription fees, low circulation and a low level of local news coverage, these papers could not attract enough advertisers and lacked appeal to the local community. Some examples include Rote Fahn (KPD), Vorw¨arts (SPD), Germania (Zentrum), Deutsche Allgemeine Zeitung (DVP) and Neue Preussische Zeitung(DNVP).

3.3.2. Mass newspapers. Competing with these elite publications for readers were mass news- papers. These papers offered light entertainment and advertisements with a strong regional focus. Berlin am Morgen (KPD) and Berliner Lokal-Anzeiger (DNVP) are some examples.

3.3.3. Tabloids. One feature that also made Berlin stand out in this period is the devel- opment of tabloid journalism. These cheap papers became very popular, especially for the left-wing and right-wing presses (Tworek, 2019). From 1925 to 1930, the Hugenberg’s rightist newspaper Nachtausgabe expanded from a circulation of 38,000 to 202,000; the M¨unzenberg leftist Welt am Abend also grew from 12,000 to 220,000. Liberal publishers like Ullstein also created several competitors: for example, the Tempo which sold 145,000 copies a day attacked from behind in the dark of night by a bow-legged, flat-footed dachshund. Bow-legged – that betrays the true race of these eastern Jewish pets, with their sagging ears and curls, who suck the marrow of our countrymen and steal the bones from under the noses of our German shepherds. Tomorrow our F¨uhrer Adolf Hitler will speak in the sports palace about this national affair, Party comrades should appear in simple battle dress, with hand grenades and flamethrowers” (Richie, 1998, pp.347-348). 12This also explains the lack of incentive to cater to public opinion in the Weimar era. Generally, the bigger a newspaper gets, the more incentive journalists have to cater to the mass. For a small newspaper, each journalist is also a member of his own audience. It is reasonable for him to believe that anything of interest to him would be equally interesting to his readers. As the newspaper gets bigger, journalists have to write for massive crowds that include different social classes and political groups. In other words, they have to write for people they do not know.

12 in 1930. The amazing array of tabloid newspapers in Berlin was the result of politics, and a symptom of the deeply divided political landscape of the Weimar Republic (Fulda, 2006). The gutter press experienced their breakthrough and existed alongside daily newspapers. Tabloids were considered more unpolitical, aiming at stimulating the readership with sen- sation. They depended much more on street sales rather than on weekly or monthly sub- scription, which also implied that they were more sensitive to popular sentiments. But it would be wrong to conclude that tabloids were apolitical. Alfred Hugenberg – leader of the German National People’s Party – asserted, “ big city folks. . . buy it because of the sensation which it carries – and they swallow the politics which is contained in between”(Holzbach, 1981). Fulda(2009) asserts, “even in self-professed ‘unpolitical’ newspapers catering to local audiences, ideological news coverage was the norm.” 13 It is worth noting that other urban and economic centers during this period – such as Hamburg, Frankfurt and Munich – did not experience a strong growth of tabloids. In 1930, the M¨unchen-Ausburger Abendzeitung blamed their low street-sales figures on the fact that Munich, in contrast to Berlin, had no proper evening rush hours (Fulda, 2006, p.189). Historians explain that the proliferation of tabloids in Berlin after 1918 was primarily driven by political concerns. Berlin was a unique political hotspot, its population of four million could accommodate a whole range of political milieux. The city also benefited from the development of a street-sale retailing structure which enabled the quick proliferation after 1918. When the Nazis took the power (1933–1945), Joseph Goebbels – the Reich Minister of Propaganda – exercised total control over the press and shut down 1000 of the 3000 news- papers.

3.4. Berlin mass transit. I investigate the development of both U-Bahn (underground railway) and S-Bahn (city rapid railway) in Berlin. After World War I, the city’s trans- portation grew rapidly, especially after the stabilization of the economy in 1924. By the end of the 1920s all the city’s railway lines were electric, and the Berlin Transport Company (BVG), uniting the tram, bus, elevated and underground train companies, was the largest community transit firm in the world (Richie, 1998, p.331). Berlin alone has 145 elevated city train stations and 70 subway stations. In 1928, the network of the elevated city trains, ring trains, and the local routes of the Reichsbahn transported 413 million people. Covering more than 55km, the subway system transported 265 million people, that is altogether 2.5 million every day (Richard Kohrherr cited in Hake(1956, p.57)). 13Gabszewicz et al.(2002) points out that softening of political opinions is not the only device used by edito- rial firms to increase their readership’s market share and thereby to please their advertisers. Diversification of their editorial content, introduction of cultural supplements, etc., are alternative methods of attracting readers.

13 3.4.1. U-Bahn. U-Bahn is the conventional rapid transit system that runs mostly under- ground. The Berlin U-Bahn was built in three major phases: 1896–1913, 1914–1930 and from 1953 onwards. I focus my attention on the second phase with the introduction of the Großprofil (large profile) network that established the first North-South line.

3.4.2. S-Bahn. S-Bahn is the hybrid of rapid transit and commuter rail that runs under- ground in the city center and above ground in the suburbs. Prior to December 1930, the Berlin S-Bahn was referred to as Berlin city, orbital and suburban railways (Berliner Stadt-, Ring- und Vorortbahnen). The Great Electrification (Groß-Elektrisierung) between 1926 and 1929, during which the core of the network was converted from steam operation to a third-rail electric railway, improved the overall service quality and capacity. In the early 1930s, the S-Bahn network continued to develop rapidly.

3.5. Newspaper distribution. The development of public transportation also led to a more widespread distribution of newspapers. The most famous and important company to deal with the sale of reading material in railway stations was Georg Stilke’s, which special- ized in books of jurisprudence and also published the important National-Liberal journal Preußische Jahrb¨ucher. In 1882, his firm obtained the concession for selling reading material in Berlin railway stations, and the business soon spread to other stations in the northern area of Germany (Reuveni, 2002). See Figure7 for one of his kiosks at Anhalter station in Berlin. At the turn of the 20th century, the number of firms distributing reading material in German railway stations was estimated at 200, after the First World War it grew to 250. According to the German Railway Authority (Reichsbahn-Gesellschaft) in 1925, more than 800 railway stations were points of distribution for reading materials. See Figure8 for a newsstand at Kaiserallee. At the time Friedrich Seidenst¨ucker took this photograph, news- stands like these had become regular features of the urban street (Magilow, 1973, p.1). These kiosks sold newspapers and magazines well suited for easy reading during the commute on Berlin’s rapidly growing and increasingly crowded subway. Rather than waiting for buyers to come to the shop, the book-peddling companies also sent their representatives to small towns and rural areas. These salesmen then went from house to house, distributing catalogues and samples of reading material (Reuveni, 2006b, p.101). In 1920, 1957 people (1119 men and 833 women) were registered in Berlin as engaging in the licensed selling of newspapers and magazines. The number rose to 3700 in 1929. The remarkable connection between public transport and newspaper consumption at- tracted a great deal of attention from contemporaries. Hale(2015) claims that in the Weimar Republic, “the number and variety of popular publications offered the traveling public by the railway newsstand as an antidote to boredom was ever a matter of astonishment to the foreign visitor.” Ernst J¨ungerdescribed in his seminal study on contemporary workers in

14 1932 how newspaper reading by workers on public transport signalled a new character of working-class life. Walther Kiaulehn, who worked for BZ am Mittag from 1930 to 1933, ex- plained that one crucial factor that contributed to the prominence of Berlin’s tabloid press was the existence of various types of public transport such as subways and buses.

3.6. Journalists in the Weimar Republic. German journalists played an exceptionally important role in Weimar society. Retallack(1993) points out, “journalists – like doctors, clerics, lawyers, and other imbued with a professional ideology – saw themselves from the outset as contributing to the good of society by both ‘enlightening’ and ‘leading’ the people”. Many journalists believed that published opinion could dictate public opinion (Tworek, 2019, p.9).14 Karl Philipp Moritz – editor of the Vossische Zeitung – commented on what he perceived as the ideal newspaper: it was no longer a vehicle for the reporting of novelties and curiosities, but rather “a mouthpiece through which one can preach to the people and force the voice of truth into both the palaces of the mighty and the hovels of the lowly.” Karl Silex – editor of the nationalist conservative newspaper Deutsche Allgemeine – explained that politicking had been “existential and inseparable from journalism” because “the generation of journalists I started with wanted to do politics” (cited in Althaus(2019)). Hale(2015) asserts, “weak on objective news coverage, as compared to Anglo-American journalism, the German press surpassed that of any other country as an educational and cultural medium. The German press was, above all others, a “serious press.” For a jour- nalist in the Weimar Republic, the winning of a mass readership was often considered less essential than selfless service to a higher ideal. Editors believed that the press carried a great educational responsibility, especially since democracy was not yet deeply rooted in German political thinking.15 Theodor Wolff – editor of the Berliner Tageblatt – explained that Germans, “whose political maturity had remained underdeveloped under its previous subordination, must be educated, above all for new tasks, a new form of government, and new responsibilities,” and the press should perform this task (S¨osemann, 2000). The his- torian and philosopher Oswald Sprengler in his best selling book The Decline of the West discusses the subject of the press: “The press today is an army with carefully organized weapons, the journalists its officers, the readers its soldiers. The reader neither knows nor is supposed to know the purposes for which he is used and the role he is to play” (Sprengler, 1928). Leydecker(2006) asserts, “certainly in no previous period of German history did

14The word “news” in English refers to “reports of recent events.” The German word for news, Nachrichten, reflects the role that news played in Germany: Nachricht emerged from the verb darnachrichten and means to direct, orient, or control something (Tworek, 2019). 15A great number of Weimar population consists of those who were mainly born in the 1880s. They reached adulthood around the turn on the century and experienced both the expansionist euphoria and the anxious forebodings of the years leading up to the First World War. The men in this age-group included many who underwent long service at the front during the war. And also because of the war, members of this generation did not gain their political experience until after 1918 (Peukert, 1987).

15 writers and intellectuals engage so directly with political events and social forces and seek so actively to have a direct influence on them as they were to do during the Weimar Republic.” As a typical broadside against journalists who wished to appear popular before the masses, “most writers who want to be popular are true beggars and toadies before the people: they stoop and submit, they try their best to affect the people’s language and flatter the people, and in particular they have to adorn their title and contents with plenty of the word ‘Volk’. Where there is real talent, the popular writer confronts the people as lecturer [Lehrer] and lord [Herr]!”. In the 1920s, most journalists continued to see themselves as educators and not so much as neutral informers (B¨osch, 2015, pp.121-122), as sages rather than purveyors of news, as intellectuals and “bearers of culture” (Kulturtr¨ager) rather than skilled technicians (Eksteins, 1975, p.73). This was also true for the popular press. A case in point is the AIZ (Arbeiter Illustrierte Zeitung) founded by the Communists, which reached a circulation of half a million copies towards the end of the Weimar era. Similar to the liberal BIZ (Berliner Illustrierte Zeitung), it printed exciting, richly illustrated reports; but the difference is that prominence was given to the oppression of the workers and the achievements of the Soviet Union. Tabloids like Hugenberg’s Berliner Lokal-Anzeiger campaigned for the Conservatives by spreading scandals about the Democrats and explicitly exhorting readers to vote for the DNVP.16 Even though the local newspapers tried, to some extent, to reflect the general political disposition of their readership, they were simultaneously a crucial factor in activating and reinforcing a political trend. Objective, factual reporting was anathema, argued one editor in a book on newspapers: News which present events in such an unclouded and unbiased fashion are only suitable for academic treatments (Fulda, 2009, p.217). Newspapers can neither use them nor do they need them. Although deploring the effects of a partisan press, German politicians largely agreed that a thoroughly objective press was undesirable as it was thought it would be too boring. There was a widespread consensus that the press must be politicized in order to fulfil its function in society.

3.7. Berlin readers. During the First World War, people already started to became more reliant on newspaper reading. This trend was reinforced during the turbulent post-war years. Newspapers provided readers with not only up-to-date information on political events, inflation, sports and the stock exchange but also entertain- ment. In 1923, in Germany’s large cities, it became a custom to spend every single free moment reading the paper. At caf´es,in doctors’ or lawyers’ waiting rooms, in trams - all that was to be seen were people’s heads buried in a pages

16While insisting on their political ideology, journalists also expressed sympathy with the audience and aimed to communicate their knowledge to the lay reader. Bowling(2012) refers to this as “empathetic expertise”.

16 of a newspaper. People became so dependent on the paper [. . . ] that many of them were afraid not to read a newspaper. (Karl B¨ucher cited in Reuveni (2006b, p.137)) As a result of politicization of the press, newspaper readers in Weimar Germany suffered from an excess of partisan information. Facts did not always help to form an objective picture of political events. “At times of political high tension, it often appears to the reader who is not hard-nosed enough to read the truth between the lines as if there were no established facts any longer. So diametrically opposed are the accounts of the situation in the papers of the various political parties”(Fulda, 2009, p.214). In May 1925, Reich Chancellor Hans Luther addressed the annual conference of the Ger- man Press Association, and proclaimed: “The German press is a Weltanschauungspresse, and it is proud of it.” The drawback of partisanship, Luther admitted, was the tendency to create small communities of faithful readers who allegedly believe everything the paper printed (Fulda, 2009, p.216).

4. Data

The data on newspapers is from the Database of Berlin daily newspapers (Datenbank der Berliner Tageszeitungen), compiled by the Austrian Academy of Sciences, which consists of 135 dailies in Berlin during the Weimar Republic. For each newspaper I observe: the circu- lation period, the predecessor, successor and any related papers (e.g., whether two papers merged, or a newspaper published a secondary edition), the frequency of publication, the name of the publisher, circulation figures17, an the political tendency or the party this news- paper supported. These 135 newspapers can be divided into six groups: elite political papers, mass papers, tabloids, foreign papers, papers for minority groups and others (unclassified). The four biggest publishers in terms of the number of daily newspapers were Ullstein, Mosse, Scherl and Germania. Supplementary data on circulation and political tendency come from de Mendelssohn(1982), Fulda(2009) and Schilling(2011). See Table1 for a summary of the data. The data on elections is from the Inter-university Consortium for Political and Social Research (ICPSR 0042 - German Weimar Republic Data, 1919 – 1933), which consists of the outcome of the 1919 election to the National Assembly and eight elections to the Reichstag. See Table2 for a complete chronological list of elections in Weimar Berlin, Figures9 and 10 for voter turnout and election results. Data for district-level elections come from B¨usch and Haus(1987). Sociodemographic and economic data at the district level come from ten issues of Statistisches Jahrbuch der Stadt Berlin published in 1924 and annually from 1926 to 1934.

17City-level circulation was decomposed into district-level circulation according to district polulation.

17 I also collect a dataset of 92 subway stations (U-Bahn) that existed in Berlin before 1933, 53 of which were opened during the Weimar era, as well as 120 city rail stations (S-Bahn), 10 of which were opened and 90 were converted from steam to third-rail electrification in the Weimar era. Using the information on current postcodes, I map each subway station to one of the 20 historical districts in 1920s.18 See Table3 for the variation in the number of railway stations and electrifications across 20 districts. Data on the annual number of tickets sold at each subway station (single ride and monthly railway tickets, tram and bus tickets) come from Gesch¨aftsbereich der Berliner Verkehrs-Aktiengesellschaft.

5. Pre-analysis

5.1. Political Tendency. According to Eksteins(1975), about one half of German news- papers in the Weimar Republic overtly acknowledged some political tendency. Those which refused to categorize themselves frequently reflected the ideas of more than one party on different political issues. For the majority of newspapers in my dataset, a specific party affil- iation is listed, or a political tendency is specified. I cross-check the data and supplement the missing information with the political tendency captured in de Mendelssohn(1982), Fulda (2009) and Schilling(2011). I also check the quality of the data by comparing to the sum- mary statistics on political orientation of the German press from 1898 to 1932 in Eksteins (1975) (see Table4).

5.2. Measure of Press Support. I use the circulation figures (the number of copies per day) as a measure of the press support. Note that in the original dataset, the circulation figures were compiled from various sources and were in general overstated, especially those from advertising catalogues (in order to impress potential advertisers), compared to those from address books, handbooks and yearbooks. The consensus among historians is that these figures had rarely been inflated by more than 30%. I use linear interpolation for the missing values, and construct an upper bound and a lower bound for the circulation figures for each year the newspaper existed. See Figure 11 for the upper bounds, lower bounds and average circulation of all newspapers in this period. The circulation figures reflected the economic conditions in the country, especially in the periods of hyperinflation (in 1923), the Golden Twenties (1924–1929) and the depression (in 1929).

5.3. Measure of Mass Polarization. The proportional representation system with mul- tiple parties representing in the parliament allows me to study the changes in mass polar- ization, which captures the divergence of the electorate’s political attitudes to ideological

18According to the Greater Berlin Act in 1920, the city was organized into 20 districts. An administrative reform in 2001 merged the existing districts into the current 12 districts. I use the information on Ortsteil of the stations to match them to the corresponding historical districts. For example, the station Breitenbachplatz now belongs to Steglitz-Zehlendorf. Since the Ortstel is Dahlem, the station belonged to the historical district Zehlendorf.

18 extremes. I construct a simple index to measure the degree of mass polarization. Following Pelizzo and Babones(2007), I compute the index polar, which measures the difference be- tween the vote shares for the extremes (left and right) and the center in each district. In the context of the Weimar Republic,

polar = (KPD + NSDAP ) − Center.

Alternatively, I compute the difference between the vote shares for the extremes and the incumbent parties in the Weimar coalition. This index ranges from −1 (total concentration) to 1 (total polarization) and captures the basic features of mass polarization in a multiple party system, essentially, “the enfeeblement of the centre, a persistent loss of votes to one of the extreme ends (or both)” (Sartori, 1976, p.136). See Table5 for a comparison of the degree of polarization between Berlin and the Weimar Republic. Notice that the index polar increased monotonically for the city of Berlin (from 0.141 to 0.641) and not for the whole country. The district Weißensee experienced the biggest upsurge in polarization, whereas Wilmersdorf experienced the smallest increase, even though the increase was more than eleven-fold (from 0.035 to 0.394).

6. Analysis

6.1. Effect of Newspaper Circulation on Election Participation and Election Out- comes. In this section, I study how changes in the circulation of newspapers affected each of the following outcome variables: voter turnout, party’s vote share and voter polarization.

6.1.1. Political Participation. Let i denote the district and t denote the election year. The outcome variable is turnoutit, the percentage of eligible voters in district i who turned out to vote in election t. The estimating equation is:

turnoutit = α + βcirit + δcontrolsit + δi + γt +εit. (1) |{z} |{z} district FE time FE

where cirit is the circulation of newspapers in district i in election t. The set of controls includes population, income19, religion (the share of Jewish and Catholic population) and employment (the share of workers in white- and blue-collar occupations) at district-election level.

In the first stage, I use the total number of stations in a district (stationit) as an instrument for the circulation of newspapers in that district:

cirit = θstationit + ρcontrolsit + νit.

19 The variable incomeit measures how relatively richer or poorer a district was compared to the aver- age income in Berlin, since income data was recorded in different currencies (Papiermark before 1923 and Reichsmark thereafter) and was also affected by inflation.

19 Table6 shows the result of this first-stage regression and Figure 12 shows the correlation between stationit (the instrument) and cirit (the endogenous variable). As expected, there is a positive relationship between the number of stations and newspaper circulation – an additional station in a district was associated with 3,653 additional copies in newspapers. F -statistic (31.57) shows that weak instrument problem is not a concern. The results are displayed in Table7.When all 20 historical districts are taken into con- sideration, an increase in circulation led to an increase in voter turnout but the effect was not statistically significant (Columns 1 – 3). Does it imply that the variation in Berlin daily newspaper circulation did not play an important role in encouraging Berliners to turn out to vote? To answer this question properly, I examine the effect across two sub-samples: the inner districts and the outer districts. The press influence was significant for 14 outer districts of Berlin: ten thousand extra copies of newspapers led to a 0.11 percentage point increase in turnout. Given that the average district population was around 200,000, this means that ten thousand additional newspapers encouraged 220 additional people in that district to turn out to vote. Formally, I use a Chow test (Chow, 1960) to check whether the coefficient estimated for the 14 outer districts is equal to the coefficient estimated for the six inner districts. Define outer as a dummy variable for outer districts. I estimate the following equation:

turnoutit = α0 + α1outer + β0cirit + β1outer × cirit

+ λ0controlsit + λ1outer × controlsit + δi + γt +εit |{z} |{z} district FE time FE and test if the estimate for β1 is zero. With p = 0.049, the result confirms the significant difference for the two sub-samples.

6.1.2. Party’s Vote Share. Next, I study the effect of the press support on partisan vote share. Let p denote parties. The outcome variable is vitp, the vote share for party p in district i for election t. The estimating equation is:

voteitp = α + βciritp + δcontrolsit + δi + γt +εitp (2) |{z} |{z} district FE time FE In the first stage, I employ two specifications. In Specification 1, I use the number of stations in district i at election t as an instrument for newspaper circulation:

ciritp = θ stationitp +ρcontrolsit + νipt (2.1) | {z } stationit∀p Note that this instrument only varies across districts and elections, and not across parties, whereas the variation of the endogenous variable ciritp is in three dimensions. This implies that for each observation with the same district and election, the same instrument is used.

20 In Specification 2, I use an interaction term between stationit and vp0 as an instrument, where vp0 is the vote share for party p at the beginning of the Weimar Republic. With this specification, the instrumental variable varies across all three dimensions.

ciript = θ IVipt +ρcontrolsit + νipt (2.2) |{z} stationit×vp0 Both specifications give similar results (see Table8). I find that an increase of ten thousand additional copies of newspapers supporting a given party led to 0.31 percentage point increase in that party’s vote share in federal elections, and 0.22 percentage point increase in Prussian state elections, but had no significant effect in Berlin city council elections. I define fed as a dummy for federal election and compare whether the influence of the daily press was statistically stronger in federal elections than in state elections by estimating the equation:

voteshareitp = α0 + α1fed + β0ciritp + β1fed × ciritp

+ λ0controlsit + λ1fed × controlsit + δi + γt +εit. |{z} |{z} district FE time FE

The Wald test rejects the null hypothesis that β1 is equal to zero at a ten percent level of significance, suggesting that the effect was statistically stronger for federal elections than for state elections.

6.1.3. Political Polarization among Voters. Another important question is how the changes in the press support affected the degree of mass polarization in Berlin. The estimating equation is

polarit = α + βcirit + δcontrolsit + δi + γt +εit. |{z} |{z} district FE time FE The first stage regression is the same as in Section 6.1.1. See Table9 for the results. I find an insignificant effect in the OLS regression after controlling for demographic variables (Column 2), but a significant effect of newspaper circulation on mass polarization in the instrument variable estimation (Columns 3 and 4). One thousand additional copies of newspapers led to an increase of 0.3943 in the degree of mass polarization. This implies that every ten percent increase in the vote shares in favor of the two extremes relative to the center can be explained by approximately 25,360 additional copies of newspapers.

6.2. Political Power of Different Types of Newspapers. In this section, I compare different types of newspapers in terms of their electoral influence by estimating equation (2) for elite political papers, and mass newspapers and tabloids separately. See Table 10 for the results.

21 With regard to how different types of newspapers benefited political parties, I find that only mass newspapers and tabloids had a significant effect on vote shares in Reichstag elec- tions, whereas the effect of elite political publications was insignificant. With regard to magnitude, the coefficient estimated for both mass newspapers and tabloids was approxi- mately twice as large as the coefficient for elite political newspapers. A Wald test on the coefficient β1 in the following equation for the pooled sample also confirms this difference:

voteshareitp = α0 + α1elite + β0ciritp + β1elite × ciritp

+ λ0controlsit + λ1elite × controlsit + δi + γt +εit, |{z} |{z} district FE time FE where elite denotes a dummy for elite political newspapers.

6.3. Political Power of the Pro- and Anti-Republication Press. Within the left- wing press, intellectuals changed swiftly between liberal, left-wing and communist newspa- pers. Kurt Tocholsky, for example, wrote for both the liberal Ullstein and the Communist M¨unzenberg publishing house. Carl von Ossietzky was initially an editor at Mosse’s Berliner Volks-Zeitung before moving on to work for the left-wing weekly journal Weltb¨uhne. Similar situation applied to the right-wing press where journalists also felt that they belonged to a wider political family. Tworek(2019) asserts that the right-wing Hugenberg’s media laid the groundwork for the Nazi government by increasing votes further on the right. This creates an incentive to examine the strength of the pro-Republican press and the anti-Republican press. I estimate equation (2) separately for the communist press (KPD), the right-wing press (DVP, DNVP and NSDAP) and the pro-Republican press (SPD, Z and DDP). I find a strong influence on the vote shares from the right-wing press but not from the communist press and the pro-Republican press. See Table 11 for the results. The influence of the right-wing press on the vote shares for its parties was three time as large as compared to the communist press’s influence on the KPD’s vote share.

7. Mechanism and Discussion

7.1. Electoral outcomes. In his pioneer work on political performance, Powell(1984) posits voter turnout, government stability and political order as the three most important indicators of democratic performance. In this paper, I find that ten thousand additional copies of newspapers in an outer district induced approximately 220 additional eligible vot- ers to turn out to vote. This can be explained by a simple argument: The availability of information increases political awareness and thus provides incentives for citizens to cast their votes. This has been studied in the literature, both theoretically (see, for example, Feddersen(2004)) and empirically. See Figure 13 for an extract from the Vorw¨arts, which targeted eligible female voters and encouraged them to vote.

22 Larcinese(2007) studies the 1997 British general election and shows that political knowl- edge and mass media play an important role in political participation. Gentzkow et al.(2011) show that entry of the first newspaper in a US county led to a small but significant increase in voter turnout. In particular, the authors find that an additional newspaper increased presidential turnout, on average, by 0.3 percentage points. The effect of the first entrant is a one percent increase in turnout, whereas second and later entrants had a significantly smaller effect on turnout. My estimate is an increase of 0.11 percentage points for each ten extra copies of newspapers. This means that if a US newspaper entered the market with a circulation of 30,000, my result is comparable to Gentzkow et al.(2011). Perlman and Schus- ter(2016) study the implementation of Rural Free Delivery in the United States in the early twentieth century and show that the increase in information to rural voters increased the vote share spread for parties but not voter turnout. They find that voter turnout decreased and newspapers circulation increased significantly over their sample period. With respect to how newspapers benefited political parties, Gentzkow et al.(2011) con- clude that partisan newspapers in the United States had no effect on vote shares. I argue that the significant effect in the Reichstag elections in the Weimar Republic can be explained by the proportional representation system, which in turn affected the voting strategy of the voters. The Weimar voters appeared to take their election duties seriously. Not only did they read a great number of newspaper articles, brochures and flyers, and study the election posters on advertising pillars (Plakats¨aulen) along the streets of Germany cities, they also participated in political meetings (Pollock, 1928). Under a majoritarian (winner-take-all) system as in the United States, what matters to a voter is whether his preferred party gets more than one half of the total votes. This means that the incentive to turn out is high if the expected margin of victory (the difference between the expected vote shares for two parties) is small. If a voter expects a party (either his preferred or less preferred party) to win by a landslide, he will be more inclined to abstain, even when his favorite newspaper encourages him to turn out and vote for its party. In other words, the electoral influence of the partisan press is reduced under the winner-take-all system. Under a proportional repre- sentation system as in the Weimar Republic, the problem of “wasted votes” is minimized, since vote shares are directly translated to seats in the parliament. The press becomes a force of greater political significance, since individual party depends more on the press as a link with its voters. To put it differently, the power sharing system helped intensify the electoral influence of the press. Other factors that may play an important role in this result are the perceptions of the press and the role of journalism in the first German democracy.20

20Eksteins(1975, p.71) explains that the Weimar press can be considered to be a force of greater political significance than a political party, since the party had to depend on the press as a link with its voters and individual politicians had relatively little contact with their constituents under the ticket-election system

23 With regard to mass political polarization, Perlman and Schuster(2016) find that the Rural Free Delivery program, which expanded the flow of information to rural areas, pro- vided a mechanism for small, populist parties to reach rural voters, and thus increased the competitiveness of congressional elections as well as the vote share for these parties. I also find that the increase in the circulation of newspapers in a district gave rise to the increase in the degree of mass polarization in that district by the same argument on competition.

7.2. Type of Newspapers. The dataset allows me to study the relative political strength of different types of newspapers. I find that it was the tabloids and mass newspapers that significantly affected election outcomes in the Weimar Republic. Elite political papers, on the other hand, had a much weaker influence. They were purchased almost exclusively by the German intelligentsia and the business community, whereas mass papers like the Berliner Morgenpost and Berliner Volkszeitung had a large working-class readership (Eksteins, 1975, p.136)21. Besides prices, an important determinant of the readership was the coverage. Layton(1970), for example, studies the V¨olkischerBeobachter – the Nazi Party’s newspaper and identifies the paper’s three political functions: to print news of the Party’s activities, to serve as an instrument of communication between Hitler and the Party rank and file, and to present the Party’s ideological views. Hitler even stated that the paper in the 1920s had been in so high an intellectual level that he himself had difficulty understanding it. Even though tabloids and mass newspapers contained more local news, entertainment and less politics than the traditional political publications, they were far from being non-partisan and had a strong influence on the voting behaviors of their readers. This is very similar to the result by Durante et al.(2019), which studies the political influence of commercial TV in Italy in the early 1980s and shows that even pure entertainment content plays a role in driving political leanings. As the authors explain, it was the language and messaging in media that affected the behavior of media consumers. For a text analysis of sensational journalism in the Weimar Republic, see Bowling(2012), which shows that sensational coverage of non-political events was heavily politicized in Weimar Berlin and argues that sensationalism provided a particularly powerful means to convey the journalist’s ideology. This idea was also supported by Weimar comtemporaries. Otto Braun – Prime Minister of Prussia from 1920 to 1932 – emphasized the need for a certain amount of sensationalism to sell politics: We like talking among ourselves condescendingly of the need for sensation of the great masses. But let us be honest: every human being has the need for a bit of sensation. The more eventful the time, the more this need becomes

(Listenwahlsystem). The role of the press was emphasized by Rudolf Kirchner – editor-in-chief of the Frank- furter Zeitung: “What effect can even the greater open-air meeting have as compared with the permanent influence a daily newspaper may exert on hundreds of thousands or even millions?” 21See Table 12 for the composition of readership for the Frankfurter Zeitung in 1927. Even though it is not a Berlin newspaper, it is comparable to other elite political newspapers in Berlin.

24 apparent, and the daily press which completely ignores this human weakness would soon appear without a reading public, because nobody goes to the newsvendor to buy sleeping pills.

7.3. Relative Strength of Left-wing and Right-wing Press. To explain the insignifi- cant effect of the communist press on the vote share of its party, I examine a survey carried out by the Communist party’s propagandists on more than 60 respondents in 1924, which attempted to answer the question: “What do workers think about Rote Fahn (the party organ)?” See Fulda(2009) for details. It is not surprising to find negative answers from the workers such as the writing style was difficult to understand, it was “more a paper for party functionaries than for the masses” and was “not entertaining enough”, it had “too many opinions and not enough news.” Indeed, the Communist newspapers, together with the Social Democratic newspapers, had the reputation to be the party press par excellence (Eksteins, 1975, pp.84-85). Their publications consisted almost solely of tedious party news and polemical attacks on opponents, and their regular readership came exclusively from the party faithful who believed strongly in the precepts of Communism or Social Democracy. In addition, it can be argued that the Communist Party put a greater emphasis on other forms of political persuasion, such as demonstrations. I collect a dataset of all protests and demonstrations from Ehls(1997), which consists of 677 protests from March 1919 to February 1933 in Berlin. For each protest, I observe the party that organized the protest, the size, the occasion and a description. Two thirds of the protests in this dataset was organized by the Communist Party. In contrast to the Communist press, the National Socialist press originated indigenously out of the fervor of local groups or individuals. Several papers were founded in depression- ridden Germany and vied for recognition as party organs. Few of them were viable as economic propositions. The purpose of the Nazi press was “to incite, to inflame, to urge on”, the effect desired was that of a “whip, which wakens the dilatory sleepers from their slumber and drives them on to indefatigable activity!” Among other major parties, the Catholic Center Party (Zentrum) and the German Na- tionalist People’s Party (DNVP) were more successful in anchoring their press to their party apparatus than were the liberal parties. For the Center Party, the bond of religion and the active interest in the press of the industrial groups within the party tied many of the Catholic papers closely to the party structure. The DNVP also had comparatively little difficulty in the 1920s in maintaining a unified front with its press partly owing to Hugenberg’s influence. The Nationalist political program during most of the Weimar Republic was essentially the opposition to the ruling coalition parties. The Nationalist press, however, frequently refused

25 to publicize differences of opinion in the right-wing camp, feeling that the cause was essen- tially the same and that only the approach of the two leaders and their parties differed. The Nationalist press thus provided a great service to Hitler. Note that the focus of my analysis is on newspapers in the city of Berlin. Provincial papers around Berlin also played an important role in the political process, but the channel through which these papers affected the voting behavior of readers was different. The density of the Berlin newspaper market was in stark contrast to the provincial press. In Berlin, readers would often be able to choose from nearly 70 daily newspapers, some of which had two or even three daily editions. Provincial readers, in contrast, were satisfied if they could enjoy a newspaper everyday of the week. As a result, newspaper readers in these provincial towns perceived a completely different kind of media tenor from those in Berlin. See Fulda(2009, Chapter4) for a study of provincial newspapers around Berlin from 1925 to 1928. Examples of these papers are the Niederbarnimer Kreisblatt, which served the reading audience in the small town Oranienburg, less than 35 kilometers from central Berlin, or the Angerm¨underZeitung, which served the town Angerm¨under,about 69 kilometers northeast of Berlin. These newspapers were typically small and suffered significantly from paper and manpower shortage.22 They easily became victims of inflation and to the Nationalists’ drive to expand their control in the press. A large fraction of the provincial papers claimed that they were renouncing their political bias for fear of a loss of advertising or subscribers, but this “purported depoliticization” usually meant a swing to the right politically. These provincial papers normally had no competitors, their news coverage remained essentially unchallenged. In the city of Berlin, however, each newspaper competed with many other papers, thus contributing to amplifying elite conflicts, magnifying and reinforcing existing political divisions.

8. Conclusion

Since the British commentator Edmund Burke pointed to the emergence of the Fourth Estate in 1787, academics, politicians and journalists have subscribed to the idea that the press or the media in general wields enormous political power. How this power actually translates into political influence and in particular voting behavior remains a challenging research question. The press apparently appears widely in different forms and serves dif- ferent purposes in different periods of history. Therefore, to properly answer this question, one should take into account the political and institutional context within which the press operates. This paper contributes to this interesting body literature by looking at the political power of the press during a unique period in history – the Weimar Republic, where the political

22See Tworek(2017) for a discussion of the paper shortage problem in the early years of the Weimar Republic.

26 system, the newspaper market structure and the journalists’ attitudes towards journalism were different from previous studies. I exploit the fact that the development in the production and distribution of the press followed to a great extent the general patterns of economic and social change, one significant element of which was the extension of the urban railway network. Different from the United States, the significant persuasive influence of the press on party’s vote shares in the Weimar era can be linked to the proportional representation system and the Weltanschauung of the correspondents. The results of this paper are in line with case studies from historians, which find that elite political publications had no effect on election outcomes. However, it does not mean that the press had no political influence. I explain the power of the press in this period by the existence of tabloids and mass newspapers, which experienced a considerable growth compared to the traditional political newspapers. It should be noted that the daily newspapers were only one kind of published texts among many that could be found in Weimar Berlin. Periodicals, magazines, pamphlets, books all made their contribution to the social and political outcomes. The Nazi party, for example, distributed millions of handbills and other publications free. But the daily newspaper press was exceptional in that it incorporated elements of all these forms of writing and engaged with its audience in a very direct way, making it a distinctive feature of Weimar society. This paper on the political power of the press seeks to make a contribution to the un- derstanding of the most “turbulent, energetic, exciting, chaotic, liberating, and frightening period in German history”.

27 References

Adena, M., R. Enikolopov, M. Petrova, V. Santarosa, and E. Zhuravskaya (2015): “Radio and the Rise of the Nazis in Prewar Germany,” The Quarterly Journal of Economics, 1885–1939. Althaus, M. (2019): “The Weimar Republic’s “Press Parliament”,” Journalism History, 44, 207–220. Bernhard, G. (1929): Der Verlag Ullstein zum Welt-Reklame-Kongress 1929, Berlin: Ull- stein, chap. Die deutsche Presse. Bosch,¨ F. (2015): Mass Media and Historical Change: Germany in International Perspec- tive, 1400 to the Present, New York and Oxford: Berghahn Books. Bowling, R. F. (2012): “Expertise and Sensational Reportage in Weimar Berlin,” Ph.D. thesis, University of Michigan. Busch,¨ O. and W. Haus (1987): Berliner Demokratie, Band I, Berlin als Hauptstadt der Weimarer Republik 1919–1933, Ver¨offentlichungen der Historischen Kommission zu Berlin, Berlin and New York: Walter de Gruyter & Co. Cage,´ J. (2017): “Media Competition, Information Provision and Political Participation: Evidence from French Local Newspapers and Elections, 1944–2014,” Journal of Public Economics, forthcoming. Chiang, C. and B. Knight (2011): “Media Bias and Influence: Evidence from Newspaper Endorsements,” Review of Economic Studies, 78, 795–820. Chiang, C.-F. (2010): “Political Differentiation in Newspaper Markets,” Unpublished. Chow, G. C. (1960): “Tests of Equality Between Sets of Coefficients in Two Linear Re- gressions,” Econometrica, 23, 591–605. Darr, J. P., M. P. Hitt, and J. L. Dunaway (2018): “Newspaper Closures Polarize Voting Behavior,” Journal of Communication, 68, 1007–1028. de Mendelssohn, P. (1982): Zeitungsstadt Berlin: Menschen und M¨achte in der Geschichte der deutschen Presse, Berlin: Verlag Ullstein GmbH. DellaVigna, S. and E. Kaplan (2007): “The Fox News Effect: Media Bias and Voting,” Quarterly Journal of Economics, 122, 1187–1234. Drago, F., T. Nannicini, and F. Sobbrio (2014): “Meet the Press: How Voters and Politicians Respond to Newspaper Entry and Exit,” American Economic Journal: Applied Economics, 6, 159–188. Durante, R. and B. Knight (2012): “Partisan Control, Media Bias, and Viewer Re- sponses: Evidence from Berlusconi’s Italy,” Journal of the European Economic Associa- tion, 10, 451–481. Durante, R., P. Pinotti, and A. Tesei (2019): “The Political Legacy of Entertainment TV,” American Economic Review, 109, 2497–2530.

28 Dussel, K. (2012): Pressebilder in der Weimarer Republik: Entgrenzung der Information, LIT Verlag M¨unster. Ehls, M. (1997): Protest und Propaganda: Demonstrationen in Berlin zur Zeit der Weimarer Republik, Die Historische Kommission zu Berlin. Eksteins, M. (1975): The Limits of Reason: The German Democratic Press and the Col- lapse of Weimar Democracy, Oxford University Press. Enikolopov, R., M. Petrova, and E. Zhuravskaya (2011): “Media and Political Persuasion: Evidence from Russia,” American Economic Review, 101, 3253–3285. Falck, O., R. Gold, and S. Heblich (2014): “E-lections: Voting Behavior and the Internet,” American Economic Review, 104, 2238–2265. Feddersen, T. J. (2004): “Rational Choice Theory and the Paradox of Not Voting,” Journal of Economic Perspectives, 18, 99–112. Fischer, E. and S. Fussel¨ (2007): Geschichte des deutschen Buchhandels im 19. und 20. Jahrhundert. Band 2: Die Weimarer Republik 1918 - 1933. Teil 2, De Gruyter, Inc. Fritzsche, P. (1996): Reading Berlin 1900, Cambridge, Massachusetts and London, Eng- land: Harvard Univeristy Press. Fulda, B. (2006): Mass Media, Culture and Society in Twentieth-Century Germany, Pal- grave Macmillan, chap. Industries of Sensationalism: German Tabloids in Weimar Berlin, 183–203. ——— (2009): Press and Politics in the Weimar Republic, Oxford University Press. Gabszewicz, J. J., D. Laussel, and N. Sonnac (2002): “Press Advertising and the Political Differentiation of Newspapers,” Journal of Public Economic Theory, 4, 317–334. Gentzkow, M. (2006): “Television and Voter Turnout,” The Quarterly Journal of Eco- nomics, 121, 931–972. Gentzkow, M. and J. M. Shapiro (2006): “Media Bias and Reputation,” Journal of Political Economy, 114, 280–316. ——— (2010): “What Drives Media Slant? Evidence from U.S. Daily Newspapers,” Econo- metrica, 78, 35–71. Gentzkow, M., J. M. Shapiro, and M. Sinkinson (2011): “The Effect of Newspaper Entry and Exit on Electoral Politics,” American Economic Review, 101, 2980–3018. ——— (2014): “Competition and Ideological Diversity: Historical Evidence from US News- papers,” American Economic Review, 104, 3073–3114. Gerber, A. S., D. Karlan, and D. Bergan (2009): “Does the Media Matter? A Field Experiment Measuring the Effect of Newspapers on Voting Behavior and Political Opinions,” American Economic Journal: Applied Economics, 1, 35–52. Groseclose, T. and J. Milyo (2005): “A Measure of Media Bias,” Quarterly Journal of Economics, 120, 1191–1237.

29 Hake, S. (1956): Topographies of Class: Modern Architecture and Mass Society in Weimar Berlin, Ann Arbor: University of Michigan Press. Hale, O. J. (2015): The Captive Press in the Third Reich, Princeton, New Jersey: Prince- ton Univeristy Press. Hamilton, J. T. (2006): All the News that’s Fit to Sell: How the Market Transforms Information into News, NJ: Princeton University Press. Haug, C. (2007): Reisen und Lesen im Zeitalter der Industrialisierung: Die Geschichte des Bahnhofs- und Verkehrsbuchhandels in Deutschland von seinen Anf¨angenum 1850 bis zum Ende der Weimarer Republik, Wiesbaden: Harrassowitz Verlag. Holzbach, H. (1981): Das System Hugenberg: Die Organisation b¨urgerlicher Samm- lungspolitik vor dem Aufstieg der NSDAP, Stuttgart: Deutsche Verlagsanstalt. Jones, A. G. (1993): Press, Politics and Society: A History of Journalism in Wales, Cardiff: University of Wales Press. Koestler, A. (1952): Arrow in the Blue, London: Collins with Hamish Hamilton. Larcinese, V. (2007): “Does Political Knowledge Increase Turnout? Evidence from the 1997 British General Election,” Public Choice, 131, 387–411. Layton, Jr., R. V. (1970): “The “V¨olkischer Beobachter,” 1920-1933: The Nazi Party Newspaper in the Weimar Era,” Central European History, 3, 353–382. Leicht, J. (2005): “Weimarer Republik: Parteien, Wahlen und Wahlrecht,” Deutsches Historisches Museum, Berlin. Lerg, W. (1989): “Media Culture of the Weimar Republic: A Historical Overview,” Journal of Communication Inquiry, 13, 94–110. Leydecker, K. (2006): German Novelists of the Weimar Republic: Intersections of Liter- ature and Politics, Boydell & Brewer, chap. Introduction, 1–17. Magilow, D. H. (1973): The Photography of Crisis: The Photo Essays of Weimar Ger- many, University Park, Pennsylvania: Pennsylvania State University Press. Mullainathan, S. and A. Schleifer (2005): “The Market for News,” American Eco- nomic Review, 95, 1031–1053. Pelizzo, R. and S. Babones (2003): “The Political Economy of Polarization: The Italian Case, 1963-1987,” Politics & Policy, 31. ——— (2007): “The Political Economy of Polarized Pluralism,” Party Politics, 13, 53–67. Perlman, E. R. and S. S. Schuster (2016): “Delivering the Vote: The Political Effect of Free Mail Delivery in Early Twentieth Century America,” Journal of Economic History, 76, 769–802. Peukert, D. J. K. (1987): The Weimar Republic: The Crisis of Classical Modernity, New York: Hill and Wang, translated by Richard Deveson.

30 Pollock, J. K. (1928): “The German Elections of 1928,” American Political Science Review, 22, 698–705. Powell, Jr., G. B. (1984): Contemporary Democracies: Participation, Stability, and Violence, Cambridge, MA: Harvard University Press. Prat, A. and D. Stromberg¨ (2013): “The Political Economy of Mass Media,” in Ad- vances in Economics and Econometrics: Theory and Applications, Proceedings of the Tenth World Congress of the Econometric Society. Preston, M. (1999): Newspapers in the Northeast: the ‘Fourth Estate’ at work in Northum- berland and Durham, Allenholme Press, chap. The Newcastle Journal 1832-1950, 112–139. Prior, M. (2013): “Media and Political Polarization,” Annual Review of Political Science, 16, 101–127. Puglisi, R. (2011): “Being The New York Times: the Political Behaviour of a Newspaper,” The B.E. Journal of Economic Analysis & Policy, 11. Retallack, J. (1993): “From Pariah to Professional? The Journalist in German Society and Politics, from the Late Enlightenment to the Rise of Hitler,” German Studies Review, 16, 175–223. Reuveni, G. (2002): “Reading Sites as Sights for Reading. The Sale of Newspapers in Germany before 1933: Bookshops in Railway Stations, Kiosks and Street Vendors,” Social History, 27, 273–287. ——— (2006a): Mass Media, Culture and Society in Twentieth-Century Germany, Palgrave Macmillan, chap. Reading, Advertising and Consumer Culture in the Weimar Period, 204–216. ——— (2006b): Reading Germany: Literature and Consumer Culture in Germany Before 1933, New York and Oxford: Berghahn Books. Richie, A. (1998): Faust’s Metropolis: A History Berlin, New York: Carroll & Graf. Sartori, G. (1976): Parties and Party Systems: A Framework for Analysis, Cambridge, England: Cambridge University Press. Schilling, K. (2011): Das Zerst¨orteErbe: Berliner Zeitungen der Weimarer Republik im Portrait, Books on Demand GmbH, Norderstedt. Snyder, Jr., J. M. and D. Stromberg¨ (2010): “Press Coverage and Political Account- ability,” Journal of Political Economy, 118, 355–408. Sobbrio, F. (2014): A Handbook of Alternative Theories of Public Economics, Edward Elgar Publishing, Inc., chap. The political economy of news media: theory, evidence and open issues, 278–320. Sosemann,¨ B. (2000): Theodor Wolff: Ein Leben mit der Zeitung, Munich: Econ Verlag. Sprengler, O. (1928): The Decline of the West, New York: Alfred A. Knopf.

31 Stromberg,¨ D. (2004): “Radio’s Impact on Public Spending,” The Quarterly Journal of Economics, 119, 189–221. Tworek, H. J. (2017): “The Death of News? The Problem of Paper in the Weimar Republic,” Central European History, 50, 328–346. ——— (2019): News from Germany: The Competition to Control World Communications, 1900-1945, Cambridge, Massachusettes and London, England: Harvard University Press. Vogel, B., D. Nohlen, and R. Schultze (1971): Wahlen in Deutschland: Theorie - Geschichte - Dokumente 1848-1970, Berlin: Walter de Gruyter. Wilke, J. (2013): “Censorship and Freedom of the Press,” In: European History Online (EGO), published by the Leibniz Institute of European History (IEG), Mainz 2013-05-08.

32 9. Appendix

Table 1. Summary Statistics of Newspapers

Types of Newspapers Elite Political Papers 20 Berlin Mass Newspapers 32 Tabloids 33 Others (Minority, Foreign, Unclassified) 50 Circulation Period Oldest Newspaper Berliner Intelligenz-Blatt (since 1826) Newest Newspaper M¨arkischesTageblatt (since Jan, 1, 1933) Longest Circulation Berliner Intelligenz-Blatt (1161 months) Shortest Circulation Die Internationale (less than one month) Average Circulation 209 months Publishers Number of Publishers 59 Largest Publishers Ullstein (7), Scherl (6), Mosse (5), Germania (5) Number of Mergers 18 Political Tendency Communist 6 Social Democrat and Left Socialist 7 Republican Democrat 7 Liberal Democrat 9 Right Democrat 1 Center (Catholic) 3 Right Conservative (National) 3 German National 9 Wirtschaftspartei 9 Right Agrarian 4 German People 2 National Socialist 5

Source: Datenbank der Berliner Zeitungen, Austrian Academy of Science

33 Table 2. Elections in Berlin in the Weimar Republic

Month–Day Year Elections Jan 19 1919 Wahl zur Verfassunggebenden Deutschen Nationalversammlung Jan 26 1919 Wahl zur Verfassunggebenden Preußischen Landesversammlung Feb 23 1919 Stadtverordnetenwahl Jun 6 1920 Reichstagswahl Jun 20 1920 Stadt- und Bezirksverordnetenwahl Feb 20 1921 Landtagswahl Oct 16 1921 Stadt- und Bezirksverordnetenwahl May 4 1924 Reichstagswahl Dec 7 1924 Reichstagswahl Dec 7 1924 Landtagswahl Mar 29 1925 Reichspr¨asidentenwahl (1. Wahlgang) Apr 26 1925 Reichspr¨asidentenwahl (2. Wahlgang) Oct 25 1925 Stadt- und Bezirksverordnetenwahl Mar 4–17 1926 Volksbegehren “Enteignung der F¨urtenverm¨ogen” Jun 20 1926 Volksbegehren “Enteignung der F¨urtenverm¨ogen” May 20 1928 Reichstagswahl May 20 1928 Landtagswahl Oct 3–16 1928 Volksbegehren “Panzerkreuzerverbot” Oct 16–29 1929 Volksbegehren “Freiheitsgesetz” Nov 17 1929 Stadt- und Bezirksverordnetenwahl Dec 22 1929 Volksentscheid “Freiheitsgesetz” Sep 14 1930 Reichstagswahl Apr 8–21 1931 Volksbegehren “Landtagsaufl¨osung” Aug 9 1931 Volksentscheid “Landtagsaufl¨osung” Mar 13 1932 Reichspr¨asidentenwahl (1. Wahlgang) Apr 10 1932 Reichspr¨asidentenwahl (2. Wahlgang) Apr 24 1932 Landtagswahl Jul 31 1932 Reichstagswahl Nov 6 1932 Reichstagswahl Mar 5 1933 Reichstagswahl Mar 5 1933 Landtagswahl Mar 12 1933 Stadt- und Bezirksverordnetenwahl Nov 12 1933 Reichstagswahl Nov 12 1933 Volksabstimmung (Austritt aus dem V¨olkerbund)

Note: I focus on federal elections (Reichstagswahl), Prussian state elections (Landtagswahl), and Berlin city council elections (Stadt- und Bezirksverordnetenwahl). Source: B¨usch and Haus(1987)

34 Table 3. New Railway Stations and Electrifications in Berlin by District and Election

New stations New electrified stations District/ Election Year 1924 1928 1930 1932 1933 1928 1930 1932 1933 Inner Districts 1. Mitte 6 1 4 1 3 2. Tiergarten 1 5 2 3. Wedding 4 2 1 4. Prenzlauer Berg 5 5. Friedrichshain 5 4 6. Kreuzberg 4 4 12 Outer Districts 7. Charlottenburg 1 3 6 9. Spandau 3 5 9. Wilmersdorf 2 3 10. Zehlendorf 3 3 3 11. Sch¨oneberg 1 1 2 3 4 12. Steglitz 1 1 13. Tempelhof 2 1 4 14. Neuk¨olln 1 4 1 2 6 15. Treptow 4 16. K¨openick 8 17. Lichtenberg 3 6 1 18. Weißensee 19. Pankow 1 2 20. Reinickendorf 11

35 Table 4. Percentage of Total Number of Newspapers in the Weimar Republic

1898 1913 1917 1926 1928 1930 1932 Right-wing 24.1 22.6 16.8 23.6 27.3 22.5 26.7 Centre Party and BVP 9.6 11.6 13.8 12.8 10.8 12.9 12.9 Left liberals 19.6 14.2 16.9 5.5 4.1 3.0 2.9 SPD and KPD 1.7 2.2 2.7 5.7 5.1 5.5 4.8 Non-partisan 45.0 49.2 49.8 52.4 52.7 55.3 52.3

Note: Table A from Appendix I in Eksteins(1975). Right-wing parties included NSDAP, DNVP, DVP and the Economic Party.

36 Table 5. Political Polarization Index

Weimar Republic Berlin

EL ER Z polarPB (%) EL ER Z polar (%) May 1924 12.6 6.6 16.6 3.6 17.9 3.8 14.1 Dec 1924 8.9 3.0 17.3 -5.4 16.3 2.0 3.9 14.4 May 1928 10.6 2.6 15.2 -2.0 24.6 1.6 3.3 22.9 Sep 1930 13.1 18.3 14.8 16.6 27.3 14.6 3.6 38.3 Jul 1932 14.6 37.3 14.2 37.7 27.3 28.7 4.9 51.1 Nov 1932 16.9 33.1 15.3 34.7 31.0 26.0 4.4 52.6 Mar 1933 12.3 43.9 14.1 42.1 24.5 34.6 5.0 64.1

Note: Columns 1 – 4 are from Pelizzo and Babones(2007). EL, ER and Z denote the vote share for the extreme left, the extreme right, and the center respectively. Note that my calculations for Berlin (in Column 8) are slightly different from theirs:

polarPB = (Communists + Nazis) − Center/BP P. BPP (Bavarian People’s Party) was included in their index since this measure was for the whole country. The index polar for the Weimar Republic did not increase monotonically, but for Berlin, it did.

37 Table 6. First-stage Regression

Newspaper Circulation (thousands) (1) (2) Number of stations 5.2287*** 3.6527*** Population (thousands) 0.5504*** Income 7.8663 Share of Jews 1.3711 Share of Catholics 6.1555*** Share of white-collar 2.7706 Share of blue-collar 3.2618*** Number of observations 286 286 F -stat 61.74 31.57 R-squared 0.1786 0.4429 ***p < 0.01, **p < 0.05, *p < 0.1

Note: All election years from 1920 to 1933 are included.

38 Table 7. Effect of Newspaper Circulation on Voter Turnout

Turnout (All Districts) Turnout (Outer Districts) (1) (2) (3) (1) (2) (3) circulation (thousands) 0.008 0.004 0.005 0.013** 0.011* 0.011* (0.012) (0.011) (0.011) (0.004) (0.004) (0.004) Demographics N Y Y N Y Y District FE N N Y N N Y Year FE N N Y N N Y Number of districts 20 20 20 14 14 14 Number of observations 280 280 280 196 196 196 R-squared 0.2439 0.1666 0.2199 0.4210 0.4112 0.4238 ***p < 0.01, **p < 0.05, *p < 0.1

Note: Standard errors are in parentheses. Each observation is a pair of district and election. All election years from 1920 to 1933 are included.

39 Table 8. Effect of Newspaper Circulation on Vote Share

Specification 1 Specification 2 All Federal State City All Federal State City circulation (thousands) 0.030** 0.029*** 0.025* 0.034 0.028** 0.031** 0.022* 0.032 Demographics Y Y Y Y Y Y Y Y District FE Y Y Y Y Y Y Y Y Year FE Y Y Y Y Y Y Y Y Party FE Y Y Y Y Y Y Y Y Observations 2069 760 740 569 2069 760 740 569 ***p < 0.01, **p < 0.05, *p < 0.1

Note: Standard errors are in parentheses. Each observation is a triple of district, election and party. All election years from 1920 to 1933 are included.

40 Table 9. Effect of Newspaper Circulation on Mass Polarization

Mass Polarization OLS IV (1) (2) (3) (4) Circulation (thousands) 0.2072*** -0.0217 0.2545*** 0.3943** Population (thousands) -0.0200 -0.1400** Income 14.6780*** -10.2112 Share of Jews -2.1007*** -3.7978*** Share of Catholics 2.5188*** 2.3301* Share of white-collar 2.5180*** 2.7549*** Share of blue-collar 1.5160*** 0.8002*** Number of observations 221 221 221 221 F -stat 198.53 189.33 R-squared 0.5759 0.8571 ***p < 0.01, **p < 0.05, *p < 0.1

Note: Standard errors are in parentheses. Each observation is a pair of district and election. All election years from 1920 to 1933 are included.

41 Table 10. Effects of Different Types of Newspapers on Vote Share

All elections Federal elections only (1) (2) (3) (1) (2) (3) cir elite (thousands) 0.017 0.018 cir mass (thousands) 0.021** 0.034** cir tabloids (thousands) 0.023** 0.038** Demographics Y Y Y Y Y Y District FE Y Y Y Y Y Y Year FE Y Y Y Y Y Y Party FE Y Y Y Y Y Y Number of observations 2069 2069 2069 760 760 760 ***p < 0.01, **p < 0.05, *p < 0.1

Note: Standard errors are in parentheses. Each observation is a tripple of district, election and party. All election years from 1920 to 1933 are included.

42 Table 11. Power of the Pro- and Anti-Republican Press

All elections Federal elections only vote L vote R vote W vote L vote R vote W cir L 0.014 0.013 (0.161) (0.211) cir R 0.033** 0.038*** (0.015) (0.011) cir W 0.021 0.024* (0.012) (0.010) Demographics Y Y Y Y Y Y District FE Y Y Y Y Y Y Year FE Y Y Y Y Y Y Number of observations 840 840 840 300 300 300

***p < 0.01, **p < 0.05, *p < 0.1

Note: Standard errors are in parentheses. Each observation is a triple of district, election and political tendency. All election years from 1920 to 1933 are included.

43 Table 12. Composition of Readership of the Frankfurter Zeitung in 1927

% Businessmen, industrialists, firms 36.9 Banks, insurance companies, and their officials 14.5 Free professions: lawyers, doctors, architects, engineers, artists, students 13.2 Hotels, clubs, libraries, restaurants 9.5 Higher civil servants 9.3 Smaller merchants, white-collar workers 6.9 Pensioners, private individuals 4.7 Others 5.0

Note: NL Payer, 17A, p. 172, reproduced in Eksteins(1975, p.129).

44 Figure 1. Extract from the Vorw¨arts(Berliner Volksblatt) on September 13, 1930

45 Figure 2. Dissemination network of newspapers

Source: Die Reklame (1932, p.193) in Reuveni(2006b, p.100)

46 Figure 3. Ullstein’s street vendor

Source: Presse- und Verlagsgeschichte im Zeichen der Eule: 125 Jahre Ullstein. Berlin: Ullstein 2002, p.54.

47 Figure 4. A Ballot Paper for Election to Reichstag in 1932

48 Figure 5. Map of Berlin after the Greater Berlin Act 1920

Source: Wittig, P. (1931): Das Verkehrswesen der Stadt Berlin und seine Vorgeschichte, Vorl¨aufigerAbdruck der ersten Abschnitte, Berlin, p. 79.

49 Figure 6. Types of Newspapers and Circulation by Year in Berlin

50 Figure 7. Kiosk at Anhalter Station

Source: Oskar H¨aring.Georg Stilke: Denkschrift und Arbeitsbericht, Berlin: Stilke 1927, p.44 in Fischer and F¨ussel(2007, p.470)

51 Figure 8. Newspaper kiosk

Source: Friedrich Seidenst¨ucker. Zeitungskiosk mit 966 Exemplaren in der Kaiserallee (Newsstand with 966 Newspapers and Illustrated Magazines in Kaiserallee, Berlin). 1932. Berlinische Galerie.

52 Figure 9. Voter turnout in Berlin

53 Figure 10. Vote shares in Berlin

54 Figure 11. Upper Bound, Lower Bound and Average Circulation by Year

55 Figure 12. District-level Number of Stations and Circulation

56 Figure 13. Extract from the Vorw¨arts(Berliner Volksblatt) on August 30, 1930

57