Thesis: This thesis consists of three empirical research papers on the political economy of China. The first chapter studies how conflict within an autocratic elite affects media content, while the second chapter shows how media content can in turn influence public opinion. The third chapter analyses the motivation and behaviour of individuals as they rise up the autocratic hierarchy.Chapter 1 offers an explanation for why media censorship varies within an autocratic country. I study how Chinese newspapers report about officials caught during Xi Jinping’s anti-corruption campaign, by collecting close to 40,000 articles in print and the corresponding social media posts and comments. I show that individuals are significantly more likely to search for and comment on news about corrupt officials from their own province. Yet, despite greater reader interest, local newspapers underreport corruption scandals involving high-level officials from their own province. Underreporting is greater when a newspaper does not rely on advertising revenue and a corrupt official is well connected. When newspapers do report on high-level corruption at home, they deemphasise these stories, by making them shorter, less negative and less likely to explicitly mention corruption. Similarly, city-level newspapers report less about corruption in their own city relative to other cities in the same province, but are more likely to report corruption within their provincial government than corresponding provincial newspapers. These results illustrate how intergovernmental conflict within an autocracy can lead to diverging media censorship strategies by different levels of government. I present suggestive evidence that this type of localised censorship can reduce the accountability of local governments.Chapter 2 investigates whether stereotypes in entertainment media promote negative sentiment against foreigners. Despite close economic ties, anti-Japanese sentiment is high in China. I assemble detailed information on Chinese TV broadcasts during 2012 and document that around 20 percent of all TV shows aired during prime time were historical TV dramas set during the Japanese occupation of China during World War II. To identify the causal effect of media on sentiment, I exploit high-frequency data and exogenous variation in the likelihood of viewing Sino-Japanese war dramas due to channel positions and substitution between similar programmes. I show that exposure to these TV shows lead to a significant increase in anti-Japanese protests and anti-Japanese hate speech on social media across China. These effects are driven by privately rather than state-produced TV shows.Chapter 3, co-authored with Paul Dutronc-Postel, illustrates how career incentives can affect bureaucrats' policy choices. We collect data on the career histories of the top bureaucrats of all Chinese prefectures between 1996 and 2014 and identify the causal effect of career incentives by exploiting variation in the ex ante competitiveness of promotions. Bureaucrats with a smaller starting cohort have a greater likelihood of promotion. This incentivises them to adopt a strategy that relies on real estate investment and rural land expropriation, resulting in faster growth in construction and GDP. We present suggestive evidence that the same incentives result in lower investment in education, public transport and health. We corroborate our findings using survey and remote sensing data, and show that land expropriations are associated with adverse outcomes for expropriated individuals, with subsequent arrests of local officials, and with the emergence of "ghost cities".