Governments use policy experiments to facilitate learning, but the nature and effects of these experiments remain unclear. We analyze China's policy experimentation since 1980—among the most systematic in history—and document three facts. First, most experiments exhibit positive sample selection. Second, local politicians exert excessive efforts during experiments that are not replicable during policies’ national rollout. Third, the central government is not fully sophisticated when interpreting experimentation outcomes. These facts suggest that policy learning may be biased and national policies may be distorted. Thus, while China’s institutions enable experimentation at an unparalleled scale, the complex political environments can also limit effective policy learning.

Online search as well as keyword-based contextual advertising on third-party publishers is primarily priced using pay-per-click (PPC): advertisers pay only when a consumer clicks on the advertisement. Slots for advertisements are auctioned, and per-click bids are weighted by the probability of a click given that the advertisement is displayed (the “click-through rate”) in addition to other factors. The PPC method allows the advertising platform (e.g. Google) to bundle together otherwise heterogeneous items (impressions on different positions on a search page, on different search phrases sharing common “keywords,” and on different publishers) into more homogeneous units, simplifying the advertiser's bidding problem. However, PPC pricing has some drawbacks. First, all clicks are not created equal: clicks on a Paris, France hotel website that is displayed on a search for Paris Hilton may result in lower profit conditional on the click. Second, for infrequently searched phrases on search engines or small content providers, it is difficult for the advertiser to accurately estimate conversion rates, increasing the risk and monitoring costs for the advertiser and diminishing their incentives to advertise broadly (indeed, on contextual networks, the advertising platform may not even provide the advertiser with sufficient accounting data about where the advertisements were displayed to allow the advertiser to distinguish sources of clicks, and the publisher mix may change on an ongoing basis.) Third, the problem of click fraud is fairly pervasive: when publishers receive a share of advertising revenue, advertisers place a single bid applying to many publishers, and revenue

Abstract Developing artificial intelligence (AI) technology requires data. In many domains, government data far exceed in magnitude and scope data collected by the private sector, and AI firms often gain access to such data when providing services to the state. We argue that such access can stimulate commercial AI innovation in part because data and trained algorithms are shareable across government and commercial uses. We gather comprehensive information on firms and public security procurement contracts in China’s facial recognition AI industry. We quantify the data accessible through contracts by measuring public security agencies’ capacity to collect surveillance video. Using a triple-differences strategy, we find that data-rich contracts, compared to data-scarce ones, lead recipient firms to develop significantly and substantially more commercial AI software. Our analysis suggests a contribution of government data to the rise of China’s facial recognition AI firms, and that states’ data collection and provision policies could shape AI innovation.

Media censorship is a hallmark of authoritarian regimes. We conduct a field experiment in China to measure the effects of providing citizens with access to an uncensored internet. We track subjects’ media consumption, beliefs regarding the media, economic beliefs, political attitudes, and behaviors over 18 months. We find four main results: (i) free access alone does not induce subjects to acquire politically sensitive information; (ii) temporary encouragement leads to a persistent increase in acquisition, indicating that demand is not permanently low; (iii) acquisition brings broad, substantial, and persistent changes to knowledge, beliefs, attitudes, and intended behaviors; and (iv) social transmission of information is statistically significant but small in magnitude. We calibrate a simple model to show that the combination of low demand for uncensored information and the moderate social transmission means China’s censorship apparatus may remain robust to a large number of citizens receiving access to an uncensored internet. (JEL C93, D72, D83, L82, L86, L88, P36)

Abstract Recent scholarship has suggested that artificial intelligence (AI) technology and autocratic regimes may be mutually reinforcing. We test for a mutually reinforcing relationship in the context of facial-recognition AI in China. To do so, we gather comprehensive data on AI firms and government procurement contracts, as well as on social unrest across China since the early 2010s. We first show that autocrats benefit from AI: local unrest leads to greater government procurement of facial-recognition AI as a new technology of political control, and increased AI procurement indeed suppresses subsequent unrest. We show that AI innovation benefits from autocrats’ suppression of unrest: the contracted AI firms innovate more both for the government and commercial markets and are more likely to export their products; noncontracted AI firms do not experience detectable negative spillovers. Taken together, these results suggest the possibility of sustained AI innovation under the Chinese regime: AI innovation entrenches the regime, and the regime’s investment in AI for political control stimulates further frontier innovation.

Abstract We show that the likelihood of a liquidity crunch in wholesale U.S. dollar funding markets depends on levels of reserve balances at the financial institutions that are the most active intermediaries of these markets. Heightened risk of an imminent liquidity crunch is signaled by significant delays in intraday payments to these large financial institutions over the prior two weeks. Our study contributes to the broader dialogue surrounding the Federal Reserve’s ongoing quantitative tightening.

Social scientists have long viewed the decision to protest as strategic, with an individual's participation a function of their beliefs about others' turnout. We conduct a framed field experiment that recalibrates individuals' beliefs about others' protest participation, in the context of Hong Kong's ongoing antiauthoritarian movement. We elicit subjects' planned participation in an upcoming protest and their prior beliefs about others' participation, in an incentivized manner. One day before the protest, we randomly provide a subset of subjects with truthful information about others' protest plans and elicit posterior beliefs about protest turnout, again in an incentivized manner. After the protest, we elicit subjects' actual participation. This allows us to identify the causal effects of positively and negatively updated beliefs about others' protest participation on subjects' own turnout. In contrast with the assumptions of many recent models of protest participation, we consistently find evidence of strategic substitutability. We provide guidance regarding plausible sources of strategic substitutability that can be incorporated into theoretical models of protests.

ABSTRACT Corporate credit lines are drawn more heavily when funding markets are stressed. This elevates expected bank funding costs. We show that credit supply is dampened by the associated debt‐overhang cost to bank shareholders. Until 2022, this impact was reduced by linking the interest paid on lines to a credit‐sensitive reference rate like the London interbank offered rate (LIBOR). We show that transition to risk‐free reference rates may exacerbate this friction. The adverse impact on credit supply is offset if drawdowns are expected to be deposited at the same bank, which happened at some of the largest banks during the global financial crisis and COVID recession.

We study the causal effect of school curricula on students? political attitudes, exploiting a major textbook reform in China between 2004 and 2010. The sharp, staggered introduction of the new curriculum across provinces allows us to identify its causal effects. We examine government documents articulating desired consequences of the reform and identify changes in textbooks reflecting these aims. A survey we conducted reveals that the reform was often successful in shaping attitudes, while evidence on behavior is mixed. Studying the new curriculum led to more positive views of China?s governance, changed views on democracy, and increased skepticism toward free markets.