Research

This paper introduces a novel method to estimate the marginal abatement cost of SO2 in Chinese manufacturing using a by-production (BP) approach. We simultaneously model desirable and undesirable outputs within a unified production system, providing insights into production technologies and firm-level efficiency in pollution abatement. To address endogeneity, we apply a two-step dynamic panel approach, identifying the marginal abatement cost from the first-order condition of profit maximization. Using firm-level production and emissions data from 1998 to 2007 in China’s chemical and non-metallic mineral industries, we find that pollution treatment facilities effectively reduce SO2 emissions. The average estimated marginal abatement cost is 6,532 yuan per ton in the chemical industry and 5,640 yuan per ton in the non-metallic mineral industry.

This paper introduces a novel by-production approach to modeling desirable and undesirable output production processes in the US banking sector. We utilize the structural proxy variable framework in which desirable outputs (different types of loans and other income-generating activities) are exogenous, which is a common practice in the banking literature. The undesirable output is non-performing loans (NPLs). To address the endogeneity of variable inputs (purchased funds and core deposits) in the production of desirable outputs, we employ an input distance function and rely on the bank’s cost-minimizing behavioral assumption. We specify the undesirable output technology as a function of desirable outputs as well as other factors such as total non-transaction accounts, undivided profits, and capital reserves. Using US commercial bank data from 2001 to 2020, we find that bank productivity exhibits steady growth in desirable outputs. Banks prioritize reducing the overall productivity impact of NPLs post-crisis, shifting focus from pre-crisis service provision.

Due to the rapid spread of the COVID-19 pandemic, accurately determining the true global infection count has become an extremely challenging task. In this context, our study explores the spatial spillover analysis of COVID-19 cases and assesses the impact of containment policy stringency on these spillovers. Furthermore, we examine the extent of under-reporting of COVID-19 cases at the country level. To account for diverse spatial dependencies, we employ a semiparametric spatial autoregressive model, in which the coefficients are smooth, unknown functions of countries’ stringency indices. Country-specific under-reporting, modeled as a one-sided deterministic function of exogenous variables, is estimated using the sieves method. Our analysis relies on COVID-19 infection data from 57 countries, which span from 2020 to 2021. We find that spillovers vary significantly across different levels of containment stringency. In addition, the true number of infections is estimated to be 1.72 to 5.73 times higher than the reported cases. These results align with previous research and have important policy implications for improving the precision of COVID-19 reporting and managing spillover effects more effectively.

During the COVID-19 pandemic, the precision in reporting infectious cases and fatalities presents significant challenges, exacerbated by rapid transmission rates and overburdened healthcare infrastructures. Officially reported cases occasionally exhibit zero increments, which is likely to be under-reported. Some models exclude zero values from the sample, creating a sample selectivity problem. In contrast, alternative models substitute zero values with a constant to enable logarithmic transformations. Since both modelling approaches are wrong, in this study, we address this issue by extending the Tobit model to account for both under-reporting and random noise. The standard Tobit model accommodates zero values but not under-reporting. Analyzing data from 61 countries between January 1, 2020, and November 3, 2020, we explore external factors that explain country-specific under-reporting. Our findings confirm the existence of under-reporting across countries and reveal that cases reported with zero increments actually involve non-zero infectious instances. This novel methodology enriches future under-reporting analyses.

This paper develops a hybrid framework to model factor-biased technical change (TC) by integrating two complementary approaches: (i) a time trend (TT) model that captures TC through temporal evolution, and (ii) a structural productivity (SP) model in which TC arises from a persistent productivity shock. The proposed framework addresses input endogeneity using a translog production function and combines both time-driven and shock-driven components to capture multidimensional productivity. This structure enables the decomposition of productivity growth into neutral and factor-biased technical change. An empirical application demonstrates the framework’s effectiveness and highlights the distinct contributions of each source to productivity dynamics.

In structural industrial organization (IO) models (Olley and Pakes, 1996; Levinsohn and Petrin, 2003; Ackerberg et al., 2015; Gandhi et al., 2020), it is typically assumed that firms allocate endogenous inputs optimally. We relax this assumption by incorporating input misallocation, thereby allowing variable input choices to deviate from their optimal levels, as defined by first-order conditions. In line with neoclassical production theory, we further distinguish between technical change (TC) and Hicks-neutral (HN) productivity. To assess the effects of misallocation, we evaluate counterfactual models that exclude input distortions. The framework is also extended to capture misallocations across two variable inputs. An empirical application using US manufacturing data illustrates the empirical relevance of our approach and highlights its implications for the estimation of productivity.