In the difference-in-differences literature, the average treatment effect on the treated (ATT) is typically considered identified under the conditional parallel trends (CPT) assumption, often without recognizing the overly restrictive nature of the covariate structure it imposes. In practice, conventional methods for assessing the CPT's plausibility rely on the assumption that covariates affect untreated potential outcomes through common coefficients across regions and periods, known as the common causality of covariates assumption (Karim and Webb 2024). Within this conventional framework, Karim and Webb (2024) demonstrate that CCC is a necessary condition for CPT. Intersection parallel trends (IPT) extend this framework by allowing more flexible incorporation of covariates, potentially revealing plausible parallel trends that traditional methods might miss. The CCC is no longer a necessary condition for intersection parallel trends (IPT), because it allows covariate effects to vary. This presentation introduces a formal statistical test for the CCC assumption, checking whether covariate coefficients are stable across regions, periods, or both. The CCC test employs a sequential F test to examine homogeneity, region-specific variation, time-specific variation, or combined variation. In the first stage, it assesses the equality of covariate coefficients across all untreated region-period cells. If rejected, the second-stage tests evaluate region and time variation separately, using Holm’s step-down procedure to control the family-wise error rate. This decision rule helps researchers choose the appropriate DID-INT model. We also provide a Stata program that automates CCC testing and suggests the best DID-INT specification. Although direct testing of parallel trends is not statistically testable, the CCC test offers guidance on selecting a specification likely to uphold intersection parallel trends, thereby yielding consistent ATT estimates.

Researchers routinely estimate costs over a lifetime or an episode, for example, expenditures after an illness. Cost data are often right-censored because of death or end of either follow-up or episode. Basu and Manning (2010) introduced an estimator that extended the existing class of two-part models to accommodate noninformative right-censoring with continuous death and censoring times. Their three-part estimator combines a parametric survival model with person-interval two-part cost models. The estimator is stable when censoring is heavy and costs are skewed with many zeros. Inverse-probability weighting methods (Bang and Tsiatis 2000; Lin 2000) become unstable in these conditions. The estimator also decomposes covariate effects on cumulative cost into effects on survival and effects on the intensity of utilization. I introduce threepm, a Stata command implementing the Basu and Manning estimator. threepm reports cumulative marginal effects and cost and decomposes between-group differences into a rate of accumulation component and a survival component. It accommodates equal or unequal interval lengths, different rates of cost accumulation during the end-of-life window, a choice of survival distributions and of cost-model families and links, and person-clustered bootstrap inference, with companion commands for specification tests, marginal effects, and plotting.

Prior research has documented associations between parenting, family, and neighborhood contexts and child behavior problems, but many findings rely on analytic approaches that are limited in their ability to account for stable family and background characteristics. Thus, the causal strength of these associations remains uncertain. This presentation examines the extent to which parenting, family, and neighborhood factors were associated with child externalizing and internalizing behavior problems using methods designed to strengthen causal inference. Using longitudinal panel data from N = 2,701 families in the future of families and child well-being study, I estimated fixed-effects, random-effects, and correlated random-effects models allowing us to compare between-family and within-family estimates while accounting for observed and unobserved time-invariant heterogeneity. Some of these models have not been previously used in analyses of parenting and child development. Fixed-effects and random-effects models are long-standing in Stata, but correlated random effects are newer to Stata. Similarities across these models are discussed. Across models, parental spanking was consistently associated with higher levels of externalizing and internalizing behavior problems, whereas increases in caregiver warmth were associated with lower internalizing behavior problems. Findings support interventions and policies aimed at reducing physical punishment and increasing parental warmth.

In India, there are many low-resource areas where pregnant women often have ≤3 ANC visits, which can lead to missed complications. Emergency medical services (EMS) transport programs were launched to help pregnant women reach healthcare facilities quickly. Research showed that these programs have not significantly reduced early neonatal deaths. To see how the number of ANC visits relates to early neonatal mortality, a secondary analysis was performed. Methods: The study included 1,431 live births from February to April 2014 among mothers who used EMS transport. Stillbirths and incomplete records were excluded during data management. Mothers' ANC visits, divided into two groups as ≤3 and ≥4, align with the minimum requirement of the WHO guidelines. Stata v19 BE is used for statistical analysis with multivariable logistic regression. The adjusted odds ratios had no significant changes in neonatal mortality on delivery method, place of delivery, or region. Results: Among mothers with minimal ANC (1–3 visits), early neonatal mortality was 7.9% (89 out of 1,128), whereas it was 1.6% (5 out of 303) among those with 4 or more visits. The unadjusted odds ratio was 5.3 (95% CI 2.1–13.2). Even after controlling for other factors, not receiving the recommended amount of antenatal care (ANC) remained associated with a higher risk.

This talk introduces a suite of new commands in Stata for financial statistics. These new commands are finreturns for easily building various types of asset returns from prices; finportfolio for constructing portfolios; finsummarize for computing summary statistics; finregress for running capital asset pricing model and Fama–MacBeth regressions; and finvalrisk for evaluating historical and model-based value at risk. I combine these commands with Stata's extensive time-series toolkit to demonstrate a workflow for analyzing financial data.

This presentation introduces npmergersim, a community-contributed command for merger simulation in differentiated-product markets when price data are unavailable, confidential, or highly aggregated. The command implements the method developed in Aguirregabiria, Clark, and Souza-Rodrigues (2026), which substitutes Bertrand pricing equations into the demand system to obtain a set of equilibrium conditions in which market shares are the only endogenous variables. This eliminates the need for price data to identify demand substitution patterns, price-cost margins, profits, consumer surplus, and merger effects. npmergersim builds directly on the mergersim command of Björnerstedt and Verboven (2014), extending it along several dimensions: estimation and counterfactual simulation using market shares alone; alternative models of firm conduct, including collusive and conduct-parameter specifications; and quantity competition as an alternative to Bertrand pricing. The command also provides a built-in specification test that compares estimates obtained with and without price data, allowing researchers to assess whether observed prices are contaminated by economically meaningful measurement error. I illustrate the command using data from the European automobile industry, replicating the merger simulation in Björnerstedt and Verboven (2014) without relying on price information.

This presentation examines how market concentration and counterfactual chain consolidation affect prices, quality, markups, and consumer welfare in Pennsylvania’s nursing home industry during 1997–2014, emphasizing differences between for-profit and nonprofit facilities. First, it investigates the relationship between concentration, private-room prices, and quality under independent-facility and chain-adjusted ownership specifications, addressing concentration endogeneity through two-step system GMM with deep-lag instruments. The estimates generally indicate a positive concentration-price relationship, while the evidence for quality remains mixed and often statistically insignificant. Alternative concentration measures and market definitions, evaluated using leave-one-out instruments, generally support this positive price relationship. Second, the study estimates standard logit and full random-coefficients demand models and recovers marginal costs and markups using Bertrand–Nash pricing conditions. The full random-coefficients model yields an average markup of $43.48 and marginal cost of $181.95, compared with $122.05 and $107.40, respectively, under standard logit. Nonprofit facilities face more price-elastic demand and have higher estimated marginal costs. Finally, using the full random-coefficients estimates, the analysis simulates chain mergers, subsequent acquisitions, and chain dissolution. Across three consolidation scenarios with homogeneous acquisitions, average prices increase by $10.18, $9.69, and $8.34 per patient day, while chain dissolution reduces prices by $5.42. Heterogeneous acquisition simulations indicate that target ownership, size, and age influence outcomes, with nonprofit targets generally generating larger price increases and consumer welfare losses than comparable for-profit targets. Overall, the findings suggest that consolidation outcomes depend on both chain ownership structure and the scale and characteristics of subsequent acquisitions.

Estimating competition using the Boone index requires the integration of multiple econometric techniques, including stochastic frontier analysis, translog cost modeling, and dynamic panel estimation. This presentation demonstrates a reproducible Stata workflow for measuring competition in the global microfinance sector using panel data from the microfinance information exchange (MIX) and the World Bank's world development indicators for the period 2010–2019. The analysis illustrates how Stata can be used to estimate a translog cost function through stochastic frontier analysis, derive institution-level marginal costs, and subsequently estimate Boone coefficients using a two-step generalized method of moments (GMM) framework. The presentation further demonstrates the use of Stata for panel-data management, model estimation, diagnostic testing, and graphical visualization of competition dynamics across countries and over time. As an empirical application, the proposed workflow is used to examine competition across microfinance institutions operating in developing economies. The results reveal predominantly negative Boone coefficients for prolonged periods in most upper and lower developing countries, indicating sustained competitive pressures, while several lower and least developed economies exhibit weaker competition as reflected by positive Boone coefficients. India demonstrates consistently strong competition throughout the study period relative to its peer economies, although an anomalous positive Boone coefficient in 2016 suggests a temporary decline in competitive intensity. Graphical analyses generated in Stata further illustrate the evolution of competition across different country groups, highlighting heterogeneous competitive dynamics associated with varying levels of economic development. Beyond its empirical findings, the presentation provides Stata users with a practical framework for implementing advanced competition analysis by integrating stochastic frontier estimation, translog cost modeling, dynamic panel GMM estimation, and reproducible visualization techniques within a single analytical workflow. The methodology is readily adaptable to other industries and countries where competition measurement using firm-level panel data is of interest.

The increasing complexity of modern datasets has heightened the demand for robust, reproducible, and efficient statistical methodologies. This presentation explores the application of Stata in addressing contemporary challenges in data management, statistical modeling, visualization, and reproducible research across diverse disciplines. I demonstrate the application of Stata for analyzing and forecasting tourism demand using modern time-series techniques. The analysis employs an ARIMA model to capture historical trends, seasonality, and temporal dependence in tourist arrivals. The modeling process includes data visualization, stationarity testing using the ADF test, model identification through ACF and PACF plots, parameter estimation, diagnostic checking, and out-of-sample forecasting. Model performance is evaluated using forecasting accuracy measures such as RMSE and MAPE. The study highlights Stata's capabilities for reproducible time-series analysis and demonstrates how ARIMA-based forecasting can provide reliable evidence to support tourism planning and decision-making under changing economic and seasonal conditions. The model ARIMA \((p,\; d,\; q)\), where \(\phi (B)(1-B)^{d}Y_{t}=\theta (B){\varepsilon }_{t}\),\(\ Y_{t}\) represents tourist arrivals at time \(t\), \(B\) is the backshift operator, \(\phi(B)\) is the autoregressive polynomial, \(\theta(B)\) is the moving average polynomial, and \({\varepsilon }_{t}\) denotes white-noise errors.

Contribute to the Stata community by sharing your feedback with StataCorp's developers. From feature improvements to bug fixes and new ways to analyze data, we want to hear how Stata can be made better for our users.