Aneja and Xu link newly-digitized personnel records of the U.S. government for 1907-1921 to census data to study the segregation of the civil service by race under President Woodrow Wilson. Using a difference-in-differences design around Wilson's inauguration, the researchers find that the introduction of employment segregation increased the black-white earnings gap by eight percentage points. This increasing gap is driven by a reallocation of existing black civil servants to lower paid positions, lowering their returns to education. Importantly, the negative effects extend beyond Wilson's presidency. Using census data for 1900-1940, Aneja and Xu show that segregation caused a relative decline in the home ownership rate of black civil servants. Moreover, by comparing children of black and white civil servants in adulthood, the researchers provide evidence that descendants of black civil servants who were exposed to Wilson's presidency exhibit lower levels of education, earnings, and social mobility. Their combined results thus document significant short and long-run costs borne by minorities during a unique episode of state-sanctioned discrimination.
In addition to the conference paper, the research was distributed as NBER Working Paper w27798, which may be a more recent version.
Fujiwara, Müller, and Schwarz study how social media affects election outcomes using variation in the number of Twitter users across U.S. counties induced by participants of the 2007 South by Southwest (SXSW) festival, a key event in Twitter's rise to popularity. They show that this variation, which remains predictive of Twitter use a decade later, is unrelated to electoral outcomes before the platform's mass adoption. Our results suggest that exposure to Twitter lowered the Republican vote share in the 2016 and 2020 presidential elections, but had limited effects on turnout and vote shares in House and Senate races, as well as previous presidential elections. Evidence from two sources of survey data indicates that the effects are driven by independent and moderate voters. Our results are consistent with the idea that Twitter's relatively liberal content can persuade voters to alter their views.
This paper documents several new facts about the Soviet Great Famine, 1932-33. There was no aggregate food shortage. Regional mortality rates were unrelated to per capita food production, but positively associated with ethnic Ukrainian population share. Political loyalty to and peasant resistance against the regime were positively associated with famine mortality and state food procurement in regions populated by ethnic Ukrainians. The findings show that, all else equal, ethnic Ukrainians suffered disproportionally high famine mortality and imply ethnic bias in famine-era policies. A back-of-the-envelope calculation indicates that ethnic bias against Ukrainians explains 77% of famine deaths in Russia, Ukraine and Belarus, and 92% in Ukraine.
This paper studies whether and how social media affect the dynamics of protests in China in the 2009-2017 period. Based on a dataset of 13.2 billion microblog posts, Qin, Stromberg, and Wu use retweets to construct a network of social media information flows across cities and exploit its rapid expansion to identify causal effects. The researchers find that despite strict government control, Chinese social media have a sizeable effect on the geographic spread of protests and strikes. While the effect is rapid and short-lived and predominantly occurs between similar events, social media considerably increase the scope of protests. Further evidence shows that the effect is likely to be driven by tacit coordination and emotional reactions rather than explicit coordination and sharing tactics. Their study sheds light on the debate regarding whether social media strengthen authoritarian regimes.
Exposing voters to non-state-provisioned information is presumed to counter incumbents' efforts to keep voters uninformed in order to remain in power. In this study, Baysan estimates the effect of randomized information campaigns on voter behavior and ideology in Turkey. The design allows her to estimate heterogeneous effects of information campaigns. Baysan finds that voter response to the same campaigns increased political polarization and the effect persisted at least two years. She concludes that reducing censorship can be polarizing and, because average measures mask both positive and negative treatment effects, the impact of information campaigns on civil society is underestimated.
In the United States, the boundaries of legislative districts are often drawn by political partisans. In the resulting partisan gerrymandering problem, a designer partitions voters into equal-sized districts with the goal of winning as many districts as possible. When the designer can perfectly predict how each individual will vote, the solution is to pack unfavorable voters into homogeneous districts and crack favorable voters across districts that each contain a bare majority of favorable voters. Wolitzky and Kolotilin study the more realistic case where the designer faces both aggregate and individual-level uncertainty, provide conditions under which appropriate generalizations of the pack and crack solution remain optimal, and analyze comparative statics. Optimal districting plans are equivalent to special cases of segregate-pair districting, a generalization of pack and crack where all sufficiently unfavorable voter types are segregated in homogeneous districts, and the remaining types are matched in a negatively assortative pattern. Methodologically, the reseachers exploit a mathematical connection between gerrymandering---partitioning voters into districts---and information design---partitioning states of the world into signals.
Many committees--juries, political task forces, etc.--spend time gathering costly information before reaching a decision. Reshidi, Lizzeri, Yariv, Chan, and Suen report results from lab experiments focused on such information collection processes. The researchers consider decisions governed by individuals and groups and compare how voting rules affect outcomes. Reshidi, Lizzeri, Yariv, Chan, and Suen also contrast static information collection, as in classical hypothesis testing, with dynamic collection, as in sequential hypothesis testing. Generally, outcomes approximate the theoretical benchmark and sequential information collection is welfare enhancing relative to static collection. Nonetheless, several important departures emerge. Static information collection is excessive, and sequential information collection is non-stationary, producing declining decision accuracies over time. Furthermore, groups using majority rule yield especially hasty and inaccurate decisions.
Under majoritarian election systems, securing participation and representation of minorities remains an open problem, made salient in the US by its history of voter suppression. One remedy recommended by the courts is Cumulative Voting (CV): each voter has as many votes as open positions and can cumulate votes on as few candidates as desired. Theory predicts that CV encourages the minority to overcome obstacles to voting: although each voter is treated equally, CV increases minority's turnout relative to the majority, and the minority's share of seats won. A lab experiment based on a costly voting design strongly supports both predictions.
In addition to the conference paper, the research was distributed as NBER Working Paper w28674, which may be a more recent version.