Search

We establish a mapping between the Chinese domestic agency ratings and S&P global ratings by matching firms' expected default probabilities (PDs) estimated using a dynamic logit model with the actual default rates of S&P ratings. The Chinese agency ratings are inflated by ten notches in light of the S&P rating standard. For example, the domestic AAA, AA, and A correspond to S&P BB+, BB, and BB- by median default probability. The PD-implied S&P ratings outperform the domestic agency ratings in predicting default and complement the latter explaining credit spread. Their superior default predictive power originates from using dynamic operating efficiency-related information. In contrast, the agency ratings give more weight to scale-based firm characteristics.

SummaryInteraction between various cognitive systems is imperative in adaptive behaviour. Yet, the possibility of concurrent cognitive control and incidental sequence learning has been contentious. To test possible interrelations, we designed a novel sequential Stroop task. As the competition theory predicted, participants’ responses followed expectations in a Stroop task, however, they did not express control-independent learning of the stimulus-control state regularities. In contrast, statistical learning of stimulus-control state associations was enhanced in high stimulus conflict. Additionally, conflict-related response cost was reduced when the high conflict trials were predictable from the sequence. Thus, cognitive conflict can benefit from incidental sequence learning. Importantly, the EEG results suggest that competition and interaction can determine different stages of information processing. It is the level of conflict, the type of sequence learning, the general task experience, and the stage of information processing which determine whether cognitive conflict and sequence learning support or compete with each other.

Background: Despite wide utilisation of severity scoring systems for case-mix determination and benchmarking in the intensive care unit, the possibility of scoring bias across ethnicities has not been examined. Recent guidelines on the use of illness severity scores to inform triage decisions for allocation of scarce resources such as mechanical ventilation during the current COVID-19 pandemic warrant examination for possible bias in these models. We investigated the performance of three severity scoring systems (APACHE IVa, OASIS, SOFA) across ethnic groups in two large ICU databases in order to identify possible ethnicity-based bias. Method: Data from the eICU Collaborative Research Database and the Medical Information Mart for Intensive Care were analysed for score performance in Asians, African Americans, Hispanics and Whites after appropriate exclusions. Discrimination and calibration were determined for all three scoring systems in all four groups. Findings: While measurements of discrimination - area under the receiver operating characteristic curve (AUROC) - were significantly different among the groups, they did not display any discernible systematic patterns of bias. In contrast, measurements of calibration - standardised mortality ratio (SMR) - indicated persistent, and in some cases significant, patterns of difference between Hispanics and African Americans versus Asians and Whites. The differences between African Americans and Whites were consistently statistically significant. While calibrations were imperfect for all groups, the scores consistently demonstrated a pattern of over-predicting mortality for African Americans and Hispanics. Interpretation: The systematic differences in calibration across ethnic groups suggest that illness severity scores reflect bias in their predictions of mortality.

This paper tests how closely the three leading market-based systemic risk measures (SRM) agree with the list of global systemically important banks (G-SIB) from the Financial Stability Board (FSB) and how closely they match the categorization of G-SIBs into the five systemic risk buckets used by the FSB to assign capital surcharges to G-SIBs. In addition, we investigate the concordance among these SRMs and with the FSB's designation methodology for G-SIBs. Finally, we test how these SRMs incorporate the information from high volatile events between 2015 to 2018. Our results show that alternative measures produce different estimates of systemic risk, systemically important banks and categories, with the SRISK ranking having the highest concordance with the FSB's classification of G-SIBs. In contrast, the three measures all promptly react to high volatile events.

We investigate the effects of attending a single-sex high school on future labor market outcomes through use of a randomized natural experiment. South Korean students are randomly assigned, by lottery, to single-sex or co-educational schools within their school districts. Using a large set of individual-level panel data, we find that entering single-sex schools significantly decreases female labor market outcomes after graduation. In contrast, we do not find evidence that single-sex education affects male outcomes. We explore possible mechanisms for these asymmetric treatment effects, including career choice and post-graduation network effects.

This paper explores the different pricing strategies of lenders who originate both government-sponsored enterprise (GSE) and non-GSE loans. We find that, conditional on loan and borrower characteristics and some observable local economic factors, mortgage rates on GSE loans vary significantly across regions. However, no sizable regional variation is observed in the loan amount or default risk. By contrast, the mortgage rates on non-GSE loans depend almost entirely on borrowers and loan characteristics. In addition, spatial variations in GSE mortgage rates are highly responsive to regional prepayment risk. The results are robust to various controls for neighborhood characteristics, including regional-level bank competition, household income, and racial composition. Overall, the findings offer a novel insight into how lenders adjust pricing strategies in response to a changing lending environment. It provides implications for the present and imminent dangers of housing bubbles predictions and the intensified refinancing wave following the COVID-19 pandemic.

Economic theory predicts that expectations on future house price growth are related to the current price of a house. We test this relationship for the supply side of the secondary housing market using micro data that links individual expectations to a subjective owner estimated value (OEV). We find a strong causal relationship that optimistic expectations indeed imply higher OEVs as compared to neutral or pessimistic expectations. We find qualitatively and quantitatively consistent results for Italy and the US as well as for booming and gloomy years. Our results survive ample robustness checks. Since we use subjective data on house prices, we first show that OEVs are indeed a valid source to study house price dynamics by performing three types of convergent validity tests. We find that price dynamics derived by either combining OEVs and dwelling characteristics, or making use of repeatedly provided OEVs by the same owner over time reproduce objectively measured market trends strikingly well – even over decades. In contrast, OEVs and objective data tend to differ in levels – potentially due to psychological bias. These results hold for a large set of countries. We hence conclude that the "wisdom of the home-owner crowd" is sufficient to study house price dynamics but OEVs are less suited for measuring the level of market prices.

The findings reported in this paper are based on analyses of a subset of data originally collected for a comparative survey of library personnel in Germany and the USA. This subset focuses on those survey respondents working in a library in the USA for a higher education institution granting doctoral degrees, and the analyses described here compare and contrast the views of those respondents holding positions as library directors with those who do not. Findings include differing views on leadership, diversity, employee morale, and meeting user expectations.

Identifying outperforming mutual funds ex-ante is a notoriously difficult task. We use machine learning to exploit fund characteristics and construct portfolios of equity funds that earn positive and significant out-of-sample alpha net of all costs. In contrast, alphas of portfolios selected with OLS are indistinguishable from zero. We show that the performance of machine-learning methods is the joint outcome of exploiting multiple fund characteristics and allowing for flexibility in the relation between characteristics and performance. Our results hold also for portfolios of only retail funds, for various measures of fund performance, for different methodological choices, and across different market conditions.

The standard metric for assessing risk in the financial realm has been the Value-at-Risk (VaR) with several parametric and non-parametric approaches and its derivatives which is Conditional Value-at-Risk (CVaR). The inability of VaR to tell loss severity beyond the confidence threshold and its incoherency gave birth to CVaR which accounted for both shortcomings and is also sub-additive. However, backtesting a 1-day CVaR model is almost impossible and VaR estimates gives better accuracy for fat tails than CVaR which makes CVaR also defective. Hence, there is need for a better measure which will capture the shortcomings of both metrics. This research will employ other risk measures beyond the conventional VaR and CVaR using the historical return of developing markets; South African Stock exchange (JTOPI-40) and the Nigerian Stock Exchange (NSE-30). In Particular, we will consider Hull-White Value-at-Risk (HWVaR) and Bubble Value-at-Risk (BVaR) and finally compare and contrast them with the two conventional metrics.

This paper derives ex-ante standard errors of risk premium predictions from neural networks (NNs). Considering standard errors, I provide improved investment strategies and ex-post out-of-sample (OOS) statistical inferences relative to existing literature. The equal-weighted (value-weighted) confident high-low strategy that takes long-short positions exclusively on stocks that have precise risk premia earns an OOS average monthly return of 3.61% (2.21%). In contrast, the conventional high-low portfolio yields 2.52% (1.48%). Existing OOS inferences do not account for ex-ante estimation uncertainty and thus are not adequate to statistically compare the OOS returns, Sharpe ratios and mean squared errors of competing trading strategies and return prediction models (e.g., linear, NN, and random forest). I develop a bootstrap procedure that delivers robust OOS inferences. The bootstrap tests reveal that large OOS return and Sharpe ratio differences between NN and benchmark linear models' traditional high-low portfolios are statistically insignificant. However, the NN-based confident high-low portfolios significantly outperform all competing strategies. Economically, standard errors reflect time-varying market uncertainty and spike after financial shocks. In the cross-section, the level and precision of risk premia are correlated, thus NN-based investments deliver more gains in the long positions.

We propose a friendship formation model that distinguishes the role of similarity and physical proximity on friendship patterns. This is a learning-based theory of friendship under which individuals spend time exploring the value of a friendship. The model predicts that friendship patterns exhibit homophily: similarity increases the likelihood of friendship. Higher proximity also increases the likelihood of friendship, and this effect is less pronounced for similar individuals: proximity fosters diversity. To verify the predictions, we use an experiment at selective boarding schools in Peru. While social networks exhibit homophily along multiple dimensions, proximity fosters more diverse friendships. This evidence stands in contrast to the predictions offered by a preference-based theory of homophily.

Organizations have recently embraced the tools of market design to address job assignment and internal mobility problems. However, matching objectives within an organization are distinct from those in other settings. We model firm-optimal job assignments and the tradeoffs of using different matching institutions. Our results illuminate where preference-respecting mechanisms (such as those developed for stable matching) perform well for organizational purposes. We then turn to a large organization's internal labor market for empirics. We find that optimal matching is highly productive. Using the organization's preferred metric, the optimal match is 36% more productive than randomly assigned matches within job categories. However, it achieves this through negative assortative matching, placing a majority of workers and managers with assignments they did not rank. By contrast, preference-based matches (using deferred acceptance) are much less productive. They are only 3% more valuable to the firm than the average random match, 24% less than the firm's optimal, and feature positive assortative matching. However, workers and managers are significantly more likely to be assigned to a preferred partner; only 25% are assigned to a match they did not rank.

Traditional models of market reactions to signals about a firm's value predict that trading volume is a V-shaped function of the signal. These models also imply that investors' posterior beliefs converge after the signal is released. However, empirical and anecdotal evidence suggest that trading volume is not V-shaped and that investors' posterior beliefs seem to diverge following the release of certain types of information. We reconcile the conflicting theoretical predictions and empirical evidence by developing a generalized model in which investors have both heterogeneous beliefs about a firm's prospects and uncertainty about the signal-precision. In contrast to traditional models, we show that some signals cause investors' beliefs to diverge and trading volume, in turn, becomes an M-shaped function of the signal. We document empirical evidence that corroborates this prediction: we show that trading volume is indeed an M-shaped function. We also develop a measure of signal-precision uncertainty and show that trading volume's M-shaped pattern is more pronounced in an environment with high signal-precision uncertainty.

How do new ideas emerge in academic contexts and what forces determine which ideas get selected and which are forgotten? We analyze more than 1,600 papers presented at the ANPEC Brazilian Economics Meetings from 2013 to 2019 using topic modeling and relative entropy measures. In contrast to simply counting citations or reference combinations, these methods explore the information in the actual texts to detect the rise of new patterns and whether these patterns persist once they have been established. We find that novelty is highly correlated with transience so that most new ideas are quickly forgotten. However, of the ideas that persist those that are more novel have a higher impact. We show that our text-based measure of impact is correlated with subsequent citations. Our results provide a metric to compare the nature of research at the level of Brazilian Economics departments as well as for individual researchers. Finally, we analyze how the selection procedures for the ANPEC meetings affect the incentives for economists to pursue more novel or conventional research.

In this paper, we study dynamic team contests. In the framework of a Tullock contest between two teams generating impacts according to the Cobb-Douglas effort aggregation function, we examine how equilibrium efforts and winning probabilities depend on the timing of the actions. We show that in contrast to synchronous contests, asynchronous contests with publicly observable actions do not result in the same equilibrium outcome as the one-stage contest; they are strategically unbalancing, leading to more lopsided contests. The results have implications about the design of team contests with complementary efforts.

We propose a new method, that we call an allocation perturbation, to derive the optimal nonlinear income tax schedules with multidimensional individual characteristics on which taxes cannot be conditioned. It is well established that, when individuals differ in terms of preferences on top of their skills, optimal marginal tax rates can be negative. In contrast, we show that with heterogeneous behavioral responses and skills, one has optimal positive marginal tax rates, under utilitarian preferences and maximin.

The COVID-19 disease has already spread to more than 213 countries and territories with infected (confirmed) cases of more than 27 million people throughout the world so far, while the numbers keep increasing. In India, this deadly disease was first detected on January 30, 2020, in a student of Kerala who returned from Wuhan. Because of India’s high population density, different cultures, and diversity, it is a good idea to have a separate analysis of each state. Hence, this paper focuses on the comprehensive analysis of the effect of COVID-19 on Indian states and Union Territories and the development of a regression model to predict the number of discharge patients and deaths in each state. The performance of the proposed prediction framework is determined by using three machine learning regression algorithms, namely Polynomial Regression (PR), Decision Tree Regression, and Random Forest (RF) Regression. The results show a comparative analysis of the states and union territories having more than 1000 cases, and the trained model is validated by testing it on further dates. The performance is evaluated using the RMSE metrics. The results show that the Polynomial Regression with an RMSE value of 0.08, shows the best performance in the prediction of the discharged patients. In contrast, in the case of prediction of deaths, Random Forest with a value of 0.14, shows a better performance than other techniques.

The economic approach to determine optimal legal policies involves maximizing a social welfare function. We propose an alternative: a consent-approach that seeks to promote consensual interactions and deter non-consensual interactions. The consent-approach does not rest upon inter-personal utility comparisons or value judgments about preferences. It does not require any additional information relative to the welfare-approach. We highlight the contrast between the welfare-approach and the consent-approach using a stylized model inspired by seminal cases of harassment and the #MeToo movement. The social welfare maximizing penalty for harassment in our model can be zero under the welfare-approach but not under the consent-approach.

New digital technologies allow companies to provide managers with performance feedback in real-time. We provide evidence on the causal effects of such real-time feedback on individual effort and the quality of initial (pre-feedback) and final (post-feedback) decisions by exploiting a natural quasi-experiment in the setting of professional soccer referees. Our results suggest that experienced agents decrease their effort under real-time feedback without harming the quality of their initial decisions. This results in better final decisions because real-time feedback allows agents to correct wrong initial decisions upon feedback. In contrast, less experienced agents increase their effort, but make worse initial decisions. Upon feedback, they manage to compensate for their increased number of wrong initial decisions, but see no improvement in the quality of final decisions. Furthermore, we document that the effects of real-time feedback change over time as agents gain more experience under this system.

The standard way to summarize the yield curve is to use the first three principal components of the yield curve, resulting in level, slope and curvature factors. Yields, however, are non-stationary. We analyze the first three principal components of yield changes, which correspond to changes in level, slope and curvature. The new factors based on changes in yields have strong predictive power for bond risk premia, in contrast to the factors based on yield levels. We also provide insights into the impact this has on the added value of macro data for bond risk premia predictions and the recent conclusion that machine learning provides better forecasts than linear regression.

We propose a family of incentive contracts that can attract some fund managers who are favored by investors and deter any manager who is unfavorable to some investors. The contract problem has hidden types, hidden actions, hidden knowledge of preferences, and opportunity cost. In contrast to standard screening contracts, our contracts neither depend explicitly on the utilities of the managers and investors nor have a menu of choices. The contracts have two crucial components: (i) a first- loss deposit to be used to offset some of the principal’s losses and (ii) a liquidation boundary. A case study is also given.

There is a long standing controversy over the magnitude of the Frisch labor supply elasticity. Macro economists using DSGE models often calibrate it to be large, while many micro data studies find it is small. Several papers attempt to reconcile the micro and macro results. We offer a new and simple explanation: Most micro studies estimate the Frisch using a 2SLS regression of hours changes on income changes. But available instruments are typically "weak." In that case, we show it is an inherent property of 2SLS that estimates of the Frisch will (spuriously) appear more precise when they are more shifted in the direction of the OLS bias, which is negative. As a result, Frisch elasticities near zero will (spuriously) appear to be precisely estimated, while large estimates will appear to be imprecise. This pattern makes it difficult for a 2SLS t-test to detect a true positive Frisch elasticity. We show how the use of a weak instrument robust hypothesis test, the Anderson-Rubin (AR) test, leads us to conclude the Frisch elasticity is large and significant in the NLSY97 data. In contrast, a conventional 2SLS t-test would lead us to conclude it is not significantly different from zero. Our application illustrates a fundamental problem with 2SLS t-tests that arises quite generally, even with strong instruments. Thus, we argue the AR test should be widely adopted in lieu of the t-test.

I propose a novel framework to quantify frequency-dependent risks in the factor zoo. Empirically, the linear stochastic discount factor (SDF) comprised of the first few low-frequency principal components (PCs) yields an out-of-sample monthly Sharpe ratio of 0.37, and other smaller low-frequency PCs are redundant. In contrast, the SDFs consisting of high-frequency and canonical PCs are dense and fail to identify slow-moving conditional information in asset returns. Moreover, I decompose the low-frequency SDF into two orthogonal priced components. The first component, linear in high-frequency PCs, is almost serially uncorrelated and relates to discount-rate news, intermediary factors, jump risk, and investor sentiment. The second component exhibits a persistent conditional dynamic and captures business-cycle risks related to consumption and GDP growth. Overall, asset pricing theory has frequency-dependent relevance.

We propose a stochastic spanning to evaluate whether anomalies are genuine under factor-model framework. Our approach is nonparametric and does not rely on any assumption of return distribution and investor risk preferences. It depends on the whole distribution of returns, rather than only on the first two moments. Of the anomalies we consider, only a few expand the opportunity set of the risk-averter and have real economic content. Our approach is consistent in identifying genuine anomalies in and out of samples. This is in contrast to mean-variance (MV) spanning tests where anomalies identified in-sample, not out-of-sample.

This paper studies the implications of disclosure repetitiveness on firm performance, information processing costs, and future stock returns. I propose a compression-based method to measure disclosure repetitiveness, which is assumption-free with respect to the underlying language models. I then decompose my measure into news repetitiveness ratio (NRR) and stale-news repetitiveness ratio (SNR), which allow me to separately analyze managers’ disclosure behaviors regarding new and old contents compared to the previous year. In contrast with previous studies, I find that operating performance and filing announcement returns are positively (negatively) correlated with NRR (SRR) and that investors under-react to information in NRR. A portfolio with long (short) positions on high (low) NRR stocks generates value-weighted alphas of 5%-7% per annum. These results are consistent with the notion that managers present good (bad) news with more (less) repetition and repeat stale news to obfuscate unfavorable information, while the investors incorporate this information slowly due to limited attention.

Deciding how to allocate the seats of a house of representatives is one of the most fundamental problems in the political organization of societies, and has been widely studied over already two centuries. The idea of proportionality is at the core of most approaches to tackle this problem, and this notion is captured by the divisor methods, such as the Jefferson/D'Hondt method. In a seminal work, Balinski and Demange extended the single-dimensional idea of divisor methods to the setting in which the seat allocation is simultaneously determined by two dimensions, and proposed the so-called biproportional apportionment method. The method, currently used in several electoral systems, is however limited to two dimensions and the question of extending it is considered to be an important problem both theoretically and in practice. In this work we initiate the study of multidimensional proportional apportionment. We first formalize a notion of multidimensional proportionality that naturally extends that of Balinski and Demange. By means of analyzing an appropriate integer linear program we are able to prove that, in contrast to the two-dimensional case, the existence of multidimensional proportional apportionments is not guaranteed and deciding its existence is NP-complete. Interestingly, our main result asserts that it is possible to find approximate multidimensional proportional apportionments that deviate from the marginals by a small amount. The proof arises through the lens of discrepancy theory, mainly inspired by the celebrated Beck-Fiala Theorem. We finally evaluate our approach by using the data from the recent 2021 Chilean Constitutional Convention election.

We introduce the new F-Riesz distribution to model tail-heterogeneity in fat-tailed covariance matrix observations. In contrast to the typical matrix-valued distributions from the econometric literature, the F-Riesz distribution allows for different tail behavior across all variables in the system. We study the consistency properties of the maximum likelihood estimator in both static and dynamic models with F- Riesz innovations using both one-step and two-step (targeting) estimation techniques. Allowing for tail-heterogeneity when modeling covariance matrices appears empirically highly relevant. When applying the new distribution to realized covariance matrices of 30 U.S. stocks over a 14 year period, we find huge likelihood increases both in-sample and out-of-sample compared to all competing distributions, including the Wishart, inverse Wishart, Riesz, inverse Riesz, and matrix-F distribution.

The mortgage market has historically been plagued by racial discrimination, and recent data show ongoing disparities in the terms offered to minority borrowers by traditional lenders. The mortgage market has changed rapidly in the past decade, however, due to the rise of technology-based (FinTech) lenders. In this paper, I show that FinTech lenders show little to no gap in the terms provided to Black and Hispanic borrowers after adjusting for GSE credit-risk pricing determinants and loan size. This fact is shown using proprietary data on borrowers to better study racial bias and FinTech loans. The paper further shows that, in a matched analysis of observationally similar Black and Hispanic and non-Black or Hispanic borrowers, FinTech lenders provide statistically indistinguishable terms to the two groups. This stands in sharp contrast to the large discrepancies found under identical tests for traditional lenders. This matched analysis, wherein borrowers with similar characteristics such as income and age are compared, is supported by an analysis of preference and shopping heterogeneity in the National Survey of Mortgage Originations. Finally, I show that FinTech penetration in a region had consequences on legacy lender terms as well. Greater FinTech penetration in a zip code is associated with smaller discrepancies even among legacy lenders.

This paper investigates the dynamic interdependence between the stock returns of regionally disjoint Japanese real estate investment trusts (REITs), where the property type and a market return are controlled. We take a multivariate time series approach with the error term being allowed to have conditional heteroscedasticity of unknown form. We find significant spillover effects from central to local areas in conditional mean, a potential signal of arbitrage opportunities. The spillover effects have become stronger after the COVID-19 crisis for the office and hotel sectors, but not for the residential sector. This contrast suggests that a geographic diversification strategy within residential REIT securities should be more effective than that within office or hotel, especially during a period of turmoil.

This paper assesses different econometric approaches to working with count-based outcome variables and other outcomes with similar distributional properties, which are increasingly common in corporate finance applications. We demonstrate that the common practice of adding a constant to the outcome and then estimating log-linear regressions produces estimates with no natural interpretation that can easily have the wrong sign. In contrast, a simple fixed-effects Poisson model produces consistent and reasonably efficient estimates under more general conditions than commonly assumed. We also show through replication of existing papers that economic conclusions are frequently sensitive to the econometric approach employed.

Using board reforms in 41 countries and staggered difference-in-differences (DiD) estimates, Fauver, Hung, Li, and Taboada (2017, FHLT) find that firm value increases following the reforms. In a study that reviews the recent econometric theory on staggered DiD estimations, Baker, Larcker, and Wang (2021, BLW) contest the robustness of FHLT’s results. They conclude that FHLT’s results become mostly insignificant using alternative approaches that address the biases suggested by this recent theory. However, we find that the insignificant findings in BLW’s analysis are largely due to selective modifications that BLW made to FHLT’s data and model specifications, which resulted in low power tests. Importantly, these changes, which were not transparently disclosed in BLW, largely explain the insignificant results reported in BLW, even before implementing the new econometric approaches. Thus, in this study, we perform a battery of tests that apply alternative approaches that address the biases suggested in this recent literature. In contrast to the conclusions in BLW, we find that FHLT’s results remain significant. Our findings reinforce the conclusion of FHLT that board reforms increase firm value and highlight the importance of fair data representation when applying modified staggered DiD methods.

We develop a flexible yet tractable model of consumer search and choice, and apply it to the problem of product rankings optimization by online retail platforms. In the model, products are characterized by an observable search index, which governs what consumers search; and a utility index, which governs which of the searched options is purchased. We show that this framework generalizes several commonly used search models. We then consider how platforms should assign products to search ranks. To optimize consumer surplus, platforms should promote “diamonds in the rough,” products whose utility index exceeds their search index. By contrast, to maximize static profits, the platform should favor high-margin products, creating a tension between the two objectives. We develop computationally tractable algorithms for estimating consumer preferences and optimizing rankings, and we provide approximate optimality guarantees in the latter case. When we apply our approach to data from Expedia, our suggested ranking achieves both higher consumer surplus and higher revenues than is achieved by the Expedia ranking, and also dominates ranking the products in order of utility.

We consider a platform that collects data from users. Data has commercial benefit to the platform, personal benefit to the user, and public benefit to other users. We ask whether the platform, or users, should have the right to decide which data the platform commercializes. We find that when users differ in their disutility from the commercialization of their data and the public benefit of data is high (low), it is welfare enhancing to let the platform (users) control the data. In contrast, when heterogeneity is in the disutility from the commercialization of different data items, it is welfare enhancing to let users (the platform) control the data when the public benefit of data is high (low).

We analyze the size and power of a large number of “robust” asset pricing tests, investigating the hypothesis that the price of risk of a candidate factor is equal to zero. Different from earlier studies, our bootstrap approach puts all tests on an equal footing and focuses on sample sizes comparable to standard applications in asset pricing research. Thus, our paper provides guidance for researchers about which method to use. We find that the classic Fama-MacBeth/Shanken approach rarely over-rejects useless factors and provides a reasonable balance between size and power. In contrast, some of the “robust” methods suffer from poor power in realistic sample sizes, especially in situations where the asset pricing model is mildly misspecified.

Can an asset manager plan the optimal timing for her/his hedging strategies given market conditions? The standard approach based on Markowitz or other more or less sophisticated financial rules aims to find the best portfolio allocation thanks to forecasted expected returns and risk but fails to fully relate market conditions to hedging strategies decision. In contrast, Deep Reinforcement Learning (DRL) can tackle this challenge by creating a dynamic dependency between market information and hedging strategies allocation decisions. In this paper, we present a realistic and augmented DRL framework that: (i) uses additional contextual information to decide an action, (ii) has a one period lag between observations and actions to account for one day lag turnover of common asset managers to rebalance their hedge, (iii) is fully tested in terms of stability and robustness thanks to a repetitive train test method called anchored walk forward training, similar in spirit to k fold cross validation for time series and (iv) allows managing leverage of our hedging strategy. Our experiment for an augmented asset manager interested in sizing and timing his hedges shows that our approach achieves superior returns and lower risk.

This paper investigates the impacts of teacher characteristics on student performance using a nationally representative and randomly assigned teacher-student sample in China. We find that having a more experienced or female homeroom teacher (HRT) with additional classroom management duties significantly improves students' test scores and cognitive and noncognitive abilities. In contrast, these effects are not observed for subject teachers who are responsible only for teaching. More experienced or female HRTs are also associated with a better classroom environment, more self-motivated students, more parental involvement, and higher parental expectations. These mechanisms explain 10-25 percent of HRT effects on test scores and cognitive ability and 50-60 percent of HRT effects on noncognitive ability. Our findings highlight the importance of teacher management skills in education production.

Despite the compelling evidence on gain-loss-domain-dependent behavior, research on domain-dependent portfolio diversification is scarce. In an online experimental study, we recruited 251 experienced US retail investors to select portfolios that differ in asset correlation and risk measures in both the gain and the loss domain. Our findings unveil that diversification behavior is domain-dependent and interacts with the level of information aggregation. Presenting aggregated portfolio outcomes prompts portfolio choices consistent with prospect theory risk preferences and generally higher levels of portfolio diversification. In contrast, displaying individual stock outcomes results in more diversified portfolios in the loss compared to the gain domain on average, in line with previous evidence suggesting that individuals do not accurately understand the concept of portfolio diversification.

The COVID-19 pandemic has created a challenging situation for everyone, sparking digital evolution due to stay-at-home restrictions to stop the spread. This has led to an uprise of digital presence, which many hypothesize has lead to a rise of cybersecurity attacks, including cyberbullying. To evaluate the significance of COVID-19 on cyberbullying reports, we collected 454, 046 of publicly available tweets from Twitter by using MongoDB and Python libraries from January 1st, 2020–June 7th, 2020. We performed statistical analyses on the collected sample set to understand the situation from a quantitative perspective. We extracted tweets related to 27 unique keywords specific to cyberbullying, including online bullying, cyberbullying, Twitter bullying, and others. Due to the time-series’ count nature, we propose a Bayesian estimation of this count data trends utilizing an autoregressive Poisson model. A simple change-point model fails to explain the subtle changes adequately. On the other hand, our Bayesian method clearly shows the upward trend beginning in mid-March, which is reportedly the time from which the stay-at-home orders were widespread globally. The pattern remains similar if we focus on one or more such keywords instead of the total count. We also provide a fine-grained lag based analysis of our model and contrast our methods with an alternative semi-Bayesian AR-ARCH model. Overall, such analysis shows somewhat conclusive evidence of the rise around the same time as COVID.

A large body of literature considers the productive advantages of cities, or “agglomeration economies”. Most empirical studies report positive agglomeration economies, although large variation exists in the magnitude of estimates. We use a meta-analysis to explore this variation, drawing on 6,684 estimates from 295 studies that cover 54 countries and span six decades. Using rich data and robust methods, we unify and extend earlier reviews. For our preferred combination of study attributes, we find agglomeration elasticities are likely to lie in the range 2.7–6.4%. Our findings confirm the controls enabled by detailed data give rise to smaller estimates. We also document several trends, with overall estimates rising from 1980–2000 and then falling. Estimates for manufacturing sectors, in contrast, fell for the entire six decades covered by our data. We speculate on possible causes of these trends, such as urban congestion, technological shocks, freight costs, and regulatory settings.

People are on the move in unprecedented numbers across the globe. How does migration affect local intergroup dynamics? In contrast to accounts that emphasize stereotypical features of groups as determinants of their treatment, we propose the social group reference dependence hypothesis: violence and negative attitudes toward each minoritized group will depend on the number and size of other minoritized groups in a community. Specifically, as groups increase in rank in their relative size (e.g., to largest minority within a community), discriminatory behavior and attitudes toward them should increase accordingly. We test this hypothesis across U.S. counties between 1990 and 2010. Consistent with this prediction we find that, as Black, Hispanic/Latinx, Asian, and Arab populations increase in rank relative to one another, they become more likely to be targeted with hate crimes and more negative attitudes. The rank effect holds above and beyond group size/proportion, growth rate, and a number of other alternative explanations. This framework makes novel predictions about how demographic shifts may affect coalitional structures in the coming years and helps explain previous findings in the literature. More broadly, our results complement the existing literature by indicating that attitudes and behaviors toward social categories are not fixed or driven only by features associated with those groups, such as stereotypes.

The authors address two significant challenges in using online text reviews to obtain fine-grained, attribute-level sentiment ratings. First, in contrast to methods that rely on word frequency, they develop a deep learning convolutional–long short-term memory hybrid model to account for language structure. The convolutional layer accounts for spatial structure (adjacent word groups or phrases), and long short-term memory accounts for the sequential structure of language (sentiment distributed and modified across nonadjacent phrases). Second, they address the problem of missing attributes in text when constructing attribute sentiment scores, as reviewers write about only a subset of attributes and remain silent on others. They develop a model-based imputation strategy using a structural model of heterogeneous rating behavior. Using Yelp restaurant review data, they show superior attribute sentiment scoring accuracy with their model. They identify three reviewer segments with different motivations: status seeking, altruism/want voice, and need to vent/praise. Surprisingly, attribute mentions in reviews are driven by the need to inform and vent/praise rather than by attribute importance. The heterogeneous model-based imputation performs better than other common imputations and, importantly, leads to managerially significant corrections in restaurant attribute ratings. More broadly, the results suggest that social science research should pay more attention to reducing measurement error in variables constructed from text.

We study the degree of synchronization between formal- and informal-economy business cycles. Using a comprehensive database of informal activity that covers a wide range of informality measures from almost 160 countries over the 1990-2018 period, we report two major results. First, fluctuations in informal-sector output are strongly positively correlated with those in formal-sector output. In contrast, fluctuations in informal employment are largely uncorrelated with those in formal-sector output. Second, movements in the formal economy tend to spillover to the informal economy. Using a novel set of instrumental variables, we show that fluctuations in formal-sector output "cause" movements in informal-sector output.

Standard choice experiments are hampered by the fact that utility is either unknown or imperfectly measured by experimenters. As a consequence, the inferences available to researchers are limited. By contrast, we design a choice experiment where the objects are valued according to only a single attribute with a continuous measure and we can observe the true preferences of subjects. Subjects have an imperfect perception of the choice objects but can improve the precision of their perception with cognitive effort. Subjects are given a choice set involving several lines of various lengths and are told to select one of them. They strive to select the longest line because they are paid an amount that increases with the length of their choice. Our design allows us to observe the search history, the response times, and make unambiguous conclusions about the optimality of choices. We find a negative relationship between the demanding nature of the choice problems and the likelihood that subjects select the optimal lines. We also find a positive relationship between the demanding nature of the choice problems and the response times. However, we find evidence that sub-optimal choices are associated with longer response times than are optimal choices. This result appears to be consistent with Fudenberg, Strack, and Strzalecki (2018). Additionally, our experimental design permits a multi-nomial discrete choice analysis. Our results suggest that the errors in our data are better described as having a Gumbel distribution rather than a normal distribution. We also observe effects consistent with memory decay and attention. Finally, we find evidence that choices in our experiment exhibit the independence from irrelevant alternatives (IIA) property.

Cross-border communication costs have plummeted and enabled the global distribution of work, but frictions attributable to distance persist. We estimate the causal effects of temporal distance stemming from time zone differences on intra-firm communication. We argue that temporal distance reduces synchronous and impromptu communication from first-best levels, and presents an especially costly friction for knowledge-intensive work. Exploiting detailed data from a large multinational firm and discrete changes in temporal distance due to shifts to and from Daylight Savings Time, we show that office pairs that lose one or two shared business hours due to DST reduce their total communication volumes by 9.2 percent on average from baseline levels, an effect driven by decreases in scheduled calls & meetings (-10.7 percent) and in instant-message chat volumes (-8.7 percent). When office pairs gain one or two shared business hours, volumes of unscheduled calls increase (+17.6 percent). Volumes of asynchronous communication (e-mail), in contrast, show no significant responses to changes in temporal distance. Examining heterogeneity by employee function, we find evidence that the shadow value of additional shared time is higher for knowledge-intensive (RD however, such adjustments do not fully offset lost communication during business hours. We discuss the implications of our findings for distributed work and for firms’ spatial organization.

Kleibergen and Zhan (Robust Inference for Consumption-based Asset Pricing, Journal of Finance, 2020) propose a new approach to testing consumption-based asset pricing models that is robust to useless factors, i.e. factors uncorrelated with the test assets. They find that recently proposed factors do not pass their test. I provide extensive simulation experiments of empirically relevant cases, which show that their method has almost no power to detect useful factors with a significant return correlation of 0.45+. In contrast, the traditional Fama-MacBeth method comes with good size and considerably higher power in small samples. Bootstrap methods can be utilized to evaluate accuracy in larger samples. I argue that the problem of useless factors is hugely exaggerated to motivate a new “robust” methodology that is actually not more “robust” than traditional methods but comes at the cost of lacking power to detect useful factors in empirically relevant cases.

Stock Connect is a stock exchange collaboration between China and international bourses such as Hong Kong and London. In contrast to existing literature mostly focusing on economic analysis, this article explores the legal issues in Stock Connect and makes practical contributions by appraising the successes and failures of this scheme.

In this paper, I study the transmission of COVID-19 in the dynamic SEIR (Susceptible, Exposed, Infectious, and Removed) model that allows individuals to optimally choose their public avoidance actions in response to the COVID-19 risk. I allow for heterogeneity in infection rates across age groups and structurally estimate the parameters to match the daily pattern of new cases and the ratio of patients by age group. Even in the absence of intervention, the elderly, who face a greater risk of death from COVID-19, are more likely than the young to take self-protective actions. In contrast to models with a fixed transmission rate, my model can capture the heterogeneity in the fraction of infected individuals among different age groups.

Women in developing countries are less likely to be entrepreneurs than men. In contrast to the previous literature on the determinants of female entrepreneurship, which looks at the role of imperfections in factor and product markets, this paper examines whether an important constraint on women starting a business is the time it takes to obtain a construction permit or operating licence, as compared to men. Using the insights of the deals and development framework, we suggest that 'de facto' deals between state and businesses, rather than 'de jure' rules, characterise the relationship that firm owners and managers have with the state. We argue that such deals are likely to be gender biased – the nature of such bias can, however, be in either direction and can differ across countries/regions too. Our analysis, based on recent rounds of World Enterprise Survey data, reveals that for most of the countries, though there is no significant difference between male- and female-owned firms in the average times taken to obtain relevant business documents, for a number of countries, the kernel densities reveal significant variation. Our econometric estimation of approximately 80 countries, however, clearly indicates gender bias of deals but the direction of bias, as a whole, is in favour of females – firms with a significant percentage of female ownership tend to require fewer days to obtain both operating licences and construction permits. Further analysis reflects distinctive patterns, based on different regions, in obtaining business-related deals.

We conduct an experiment in which subjects play an ultimatum game but, rather than bargaining over money, they bargain over lottery tickets for a prize. Compared to the standard ultimatum game, proposers offer a significantly lower percentage of lottery tickets, which is inconsistent with either ex ante or ex post fairness. In contrast, responders have a significantly higher acceptance threshold, which is consistent with ex post fairness. By varying the timing of the accept/reject decision of responders, we also show that intentions matter and present evidence of a choice anomaly in responder preferences concerning their willingness to accept extreme inequality.