To make the conventional synthetic control methods more flexible to estimate the average treatment effect (ATE), this presentation proposes a quasisynthetic control method for nonlinear models under the index model framework with possible high-dimensional covariates. The presentation suggests using the minimum average variance estimation (MAVE) method to estimate parameters and the lasso-type procedure to choose high-dimensional covariates. We derive the asymptotic distribution of the proposed ATE estimators for both finite and diverging dimensions of covariates.

This presentation will demonstrate how to produce informative, robust, and complex graphs using reproducible official and community-contributed routines in Stata. We will also discuss commonly used programming tools and tips for creating more engaging graphs.

How to solve the negative impact of environmental pollution caused by energy consumption on public health is an important challenge to achieving the goal of a healthy China, and the development of clean energy provides a feasible governance path for this. This presentation takes the commissioning and operation of the West–East Gas Pipeline II Project as a quasinatural experiment and uses the China Health and Nutrition Survey (CHNS) data from 2006 to 2015 to empirically examine how clean energy development affects public health. The study found that the West–East Gas Pipeline Project has produced health effects, and after passing multiple robustness tests, it can still significantly improve the public health level in the areas along the route. However, this effect is mainly reflected in urban residents and the elderly, and the improvement of household energy consumption structure, enterprise pollution reduction, and improvement of urban environmental quality are the main channels of action. Further analysis shows that the “coal to gas” policy helps to enhance the health effects of the project.

This presentation focuses on the current status of new quality productivity research and the issues that should be paid attention to in the measurement and evaluation process. No substantial calculation of the new quality productivity development index has been carried out. It is recommended that the National Bureau of Statistics should regularly produce and publish the "New Quality Productivity Development Index" as a product, rather than having it published by the private sector.

Average causal response function (ACRF) is a useful tool to assess treatment effect with dose functions, especially when the treatment is endogenous. This presentation presents the identification and estimation of an ACRF with sample selection and high-dimensional controls. We derive the Neyman-orthogonal moments with multiple nuisance parameters.

As one of the most important quasiexperimental causal inference methods, regression discontinuity design has two major analytical frameworks, which are quite different in terms of premise assumptions, bandwidth selection, and inference methods. Among them, the continuity-based framework assumes that the conditional expectation of the potential results is continuous and is widely used in empirical research. The local randomization framework is a rising star. This framework assumes that the driving variables can be regarded as randomly assigned in a small window near the breakpoint. This presentation will introduce the principles and techniques of these two frameworks, including identification, estimation, and inference, and compare the differences between the two through Monte Carlo simulation and Stata cases, as well as their application prospects.

The interaction effect in endogenous probit models with an interaction term is consistently estimated in Zhou and Li (2021). However, the estimation and test are time-consuming when the sample size is large. In this presentation, a new Stata command, eivprobit, is developed to implement Zhou–Li's method in much less time. Besides, the marginal effects of the two interacted regressors and the quadratic effect of a regressor with a squared term can also be estimated by the command. The eivprobit estimation is based on the control function approach and the standard errors of the estimated effects are obtained by nonparametric bootstrapping. Moreover, the finite sample Monto Carlo simulation shows that the estimator of the interaction effect behaves well and better than the usual methods such as Ai and Norton (2003)'s estimator ignoring endogeneity or the coefficient estimator of the interaction term in IV-probit estimation.

Mediation models with censored outcomes play a crucial role in social and medical sciences. However, the inherent censoring characteristics of the data often lead existing models to rely on assumptions of linearity, homogeneity, and normality for estimation. Unfortunately, these assumptions may not align with the complexities of real-world problems, limiting the persuasiveness of causal analyses. In this study, I investigate causal mediation analysis within a counterfactual framework by framing it as a neural style transfer problem commonly encountered in image processing. Acknowledging the impressive capabilities of generative adversarial networks (GANs) in handling neural style transfer, I propose a novel GAN-based model named generative adversarial censored mediation network to address mediation issues under my concern. My model employs rectified linear unit (ReLU) activation function and designs a particular multichannel network structure to implement the censored outcome mechanism while accommodating multiple mediators. To guide my model in accurately learning the underlying data patterns, I also develop a novel min-max optimization problem. Leveraging the strengths of GANs, my model fundamentally relaxes the stringent assumptions present in traditional models, resulting in more precise estimations of mediation effects and promising inference outcomes, especially in the context of intricate data patterns. Through unique insights and techniques, this study illustrates how generative learning methods can serve as an effective and robust approach for diverse causal mediation problems. I substantiate our claims with numerical results obtained from synthetic and realistic datasets, showcasing the superior performance of my method.

You can use treatment-effects estimators to draw causal inferences from observational data. You can use lasso when you want to control for many potential covariates. With standard treatment-effects models, there is an intrinsic conflict between two required assumptions. The conditional independence assumption is likely to be satisfied with many variables in the model, while the overlap assumption is likely to be satisfied with fewer variables in the model.

Over the past thirty years, project evaluation econometrics has experienced great development. Thanks to the continuous progress of econometric analysis software technology and the increasing abundance of data, the empirical research paradigm of economics and even the research paradigm of economics as a whole has undergone a huge transformation, which has profoundly affected the teaching and research of economics. In the past five years, mainstream project evaluation econometric methods such as DID, IV, and RD and a number of new econometric theoretical advances have emerged. On the one hand, they have repaired and improved the original theoretical methods and application practices, and on the other hand, they have promoted the further development of project evaluation econometric methods. I will give a general introduction to these advances.

The peer-effect (or neighbor-effect) model is an important model for studying the mutual influence between individuals. Its setting is similar to that of the spatial econometric model. However, in the spatial econometric model, the adjacency matrix is often regarded as exogenous. If the adjacency matrix is not a geographical network, but a social or economic network then the exogeneity assumption is unreasonable. This presentation proposes Stata estimation commands for heterogeneous peer-effect models and endogenous peer-effect models, snreghnet and snregenet. snreghnet can examine the row heterogeneity and column heterogeneity of the model. snregenet calculates the two-stage instrumental-variables estimation of the model and uses wild bootstrapping to calculate the standard error.

Li et al. (2024) extended the factor-based approach for estimating and inferring treatment effects in interactive fixed-effect panel models. This presentation introduces a new Stata command, xtteifeci, which generates confidence intervals and p-values for treatment effects period by period and supports a variety of model settings, including models that include covariates and nonstationary trends. Finally, the specific operation of the command is introduced in detail with a classic case.

With the rapid development of information technology and the deepening of global economic integration, the digital economy has become an important engine for driving global economic growth. However, due to differences in resource endowments, economic foundations, policy support and other factors in different regions, the development level of the digital economy shows obvious spatial heterogeneity in different regions. Therefore, it is necessary and of practical significance to conduct an in-depth analysis of the spatial heterogeneity of the development level of the digital economy. This will not only help us fully understand the distribution and development trends of the digital economy in the country and even the world, but also help to reveal the interactive relationship between the digital economy and regional economic development, identify the development gaps and potential risks of the digital economy between regions, and promote the sustainable development of the economy.

With the development of the times and the advancement of technology, general statistical data has been widely used. At the same time, unstructured data in the form of text is gradually becoming the backbone of the empirical field of business management. It is worth noting that researchers usually give priority to using tools such as Python when conducting text analysis. However, the migration of tools from Stata to Python is often accompanied by a considerable learning cost. In this case, we can't help but wonder, can we use Stata to do text analysis content? This topic aims to introduce the mainstream text analysis methods in the field of business management, and explore the possibilities and limitations of using Stata for these text analyses.

In the case of perturbations when a treatment of one unit also affects the outcomes of other units, the SUTVA assumption of traditional causal inference is violated. When perturbations work, policy evaluation relies mainly on the assumptions of randomized experiments under cluster perturbations and binary treatments. Instead, we consider non-experimental treatments under continuous treatments and network perturbations. Specifically, we define spillover effects by defining the exposure to network treatments as the weighted average of the treatments received by units connected by physical, social, or economic interactions. Building on Forastiere et al. (2021), we provide an estimator based on generalized propensity scores to estimate the direct and spillover effects of continuous treatments. Our estimator also allows for the consideration of asymmetric network connections characterized by different strengths. This presentation introduces a new Stata command that combines the advantages of Mathematica, estimates individual propensity scores and neighborhood propensity scores using linear regression and machine learning methods, supports multiple model settings, estimates the outcome model and dose-response function (ADRF) using generalized linear models, and calculates standard errors using bootstrapping. Finally, a practical example is used to introduce the specific operation of the command in detail.

Stata's official commands merge and joinby can perform horizontal merges of 1:1, m:1, 1:m, and m:m data. However, these two commands require the use of datasets to be saved to the hard disk, which not only increases the time cost but also may generate a large number of intermediate files, affecting efficiency. Stata's data frame function allows users to operate multiple datasets in memory at the same time without saving the data to the hard disk but its frlink and frget commands support only 1:1 and m:1 merge types, not 1:m and m:m. The usual solution is to save the data using the data frame to the hard disk, and then merge the data into the main data frame with the help of the merge and joinby commands.

Obviously, when the amount of data to be merged is large or there are many datasets involved, this method is very inefficient. This talk introduces a new command, framerge, to solve the above problems. The framerge command not only supports multiple data merge relationships (including 1:1, m:1, 1:m, and m:m) but can also directly operate data in memory without reading and writing hard disks, thereby improving the efficiency of data processing. The specific operation of the command is introduced in detail with a case. Propensity score supports multiple model settings, uses generalized linear model to fit the outcome model and dose–response function (ADRF), and uses bootstrapping methods to calculate standard error.