Zihua (James) LIU

We examine whether and how variations in country-level institutional factors explain the intensity, diversity and informativeness of corporate disclosures around the world. Using a comprehensive corporate disclosure dataset containing more than 100 types of disclosures from firms domiciled in 35 countries, we examine the effect of four core country-level institutional factors—legal system, creditor/investor rights, political process and societal characteristics—on corporate disclosures. Our results suggest that the country-level institutional factor, which is likely to capture the legal system of a country, is negatively associated with the intensity, diversity and informativeness of disclosure. Moreover, our results suggest that the level of creditor/investor rights protection, political process and societal characteristics can also consistently affect the production, diversity and informativeness of disclosures. Overall, our evidence broadens our collective understanding of how core institutional factors at the society and country levels systematically explain corporate disclosures and their associated informativeness.

We study the effect of political power on resource allocation for knowledge production dictated by central planning in a non-market system. Our empirical results suggest that, compared to non-connected scholars, political connected (PC) scholars have 15.7% more allocation granted by the national Natural Science Foundation of China (NSFC). Variations in grant allocation is related to weaker institutional environments, less reputable universities, and hard-to-value project. Additional analysis suggests that access to the NSFC fund not only benefits individual PC scholars in terms of their research quality, but also brings more high-impact publications for their affiliated institution.

We examine how the level of negativity of US financial analysts’ mother tongue language affects their earnings forecast optimism. By collecting negative emotional words (expressing death, sadness, diseases and violence) from 24 different languages, we construct a novel measure of language negativity at the country-level for 47 countries, capturing a country's general tendency to use negative narratives in citizen’s daily life. We find that financial analysts with their mother tongue language characterized by a high level of negativity tend to issue less optimistic earnings forecasts. Additional result suggests that the effect of language negativity on analyst forecast optimism tends to be stronger (1) during financial crisis period when market sentiment is low; (2) for firms with loss and a high level of earnings volatility, and (3) for younger analysts and analysts working for a smaller broker. The research results shed light on how language styles may shape professional financial participant behavior and lead to market impact. Overall, the findings of our study support the conjecture that the level of narrative negativity across languages can have a significant impact on capital market participants’ behavior.

TBA;

A landscape of a complex system presents a quantitative measure of its global state. The profile of research strength in Mainland China is investigated in detail, by which we illustrate a complex network based framework to extract a landscape from detailed records. First, a measure analogous to the Jaccard similarity is proposed to calculate from the presided funds similarities between the top-ranked universities. The neighbor threshold method is employed to reconstruct the similarity network of the universities. Second, the network is divided into communities. In each community the node with the largest degree and the smallest average shortest path length is taken as the representative of the community, called central node. The node bridging each pair of communities is defined to be a boundary. The central nodes and boundaries cooperatively give us a picture of the research strength landscape. Third, the evolutionary behavior is monitored by the fission and fusion probability matrices, elements of which are the percentage of a community at present time that joins into every community at the next time, and the percentage of a community at next time that comes from every present community, respectively. The landscapes in three successive 4-year durations are identified. It was found that some types of universities, such as the medicine&pharmacy and the finance&economy, conserve in single communities in the more than ten years, respectively. The agriculture&forest universities tend to cluster into one community. Meanwhile the engineering type distributes in different communities and tends to mix with the comprehension type. This framework can be used straightforwardly to analyze temporal networks. It provides also a new network-based method for multivariate time series analysis.