This paper contributes to the discussion of public and private investment in Brazil by analyzing how internal and external shocks are transmitted to changes in the capital stock. The baseline specification is estimated using a Time-Varying Parameter VAR with dynamic structures, following Gabauer and Gupta (2018). The aim is to measure the magnitude and direction of pairwise spillovers, accounting for their persistence across different time horizons. Results from April 2007 to September 2024 indicate that changes in the capital stock mostly serve as net shock receivers. Most of this shock transmission occurs at frequencies longer than one year. This study also highlights the specific link between changes in the capital stock and economic growth. To validate the main findings, this paper employs sensitivity tests on the forecast horizon, the lag order, the forgetting factors, the frequency partition, and a 200-draw randomization over the hyperparameters. Based on additional tests, the proposed model appears stable across forecast horizons, as evidenced by the TCI's low variability, whereas the pairwise results involving the SELIC rate and inflation should be viewed with greater caution, as they are less robust.
The synthetic control method (SCM) estimates the treatment effect for a single treated unit by constructing a counterfactual outcome as a weighted average of the outcomes from untreated donor units. SCM requires that the estimated weights be greater than zero and sum to one. However, these constraints can be suboptimal if the treated unit's time series exhibits consistently higher or lower values or variance, or opposite trends relative to the donor units. To address this limitation, we systematically evaluate a regularized version of SCM (RSCM), allowing weights to take on any value. Simulation results show that SCM and RSCM perform equally well when outcomes of the treated unit are located inside the outcome set of donor units. However, when treated unit outcomes are at the boundary, RSCM outperforms SCM. Finally, placebo tests in two empirical applications confirm that RSCM accurately predicts employment levels for different labour market sectors.
This study empirically investigates the impact of COVID-19 severity on the quarterly performance of Japanese overseas affiliates from 2020 to 2022, focusing primarily on sales. In particular, we highlight the role of intra-firm networks, specifically the presence of nearby affiliates under the same parent company, in mitigating the adverse effects. We find that higher local COVID-19 severity significantly reduced total sales, with the negative impact being most pronounced during the initial phase; conversely, Japanese affiliates rarely reduced employment. We also find that, on average, intra-firm networks within the same region did not necessarily mitigate the negative effects of local pandemic severity. However, we detected such mitigating effects among affiliates in East Asia with sibling affiliates in less-affected countries of the region during the initial year. This was largely because they could more easily benefit from the positive aspects of network effects (e.g., switching input sources) within the dense production networks spanning multiple countries in the region, taking advantage of variations in the timing and level of pandemic severity across those countries. Additionally, affiliates in East Asia experienced negative supply chain spillovers from the pandemic’s severity in input-source countries during the initial year, reflecting the extensive cross-border nature of these chains.
This article examines how thresholds in tax relief policies can create incentives for businesses to strategically sort around eligibility cut-offs. We study Scotland’s Small Business Bonus Scheme, which provides full relief from business rates for non-domestic properties with rateable value below a fixed threshold, and partial relief above it. Using administrative data for the city of Glasgow (2009–2019), we estimate a regression discontinuity around the eligibility cut-off and find a higher concentration of active businesses below the 100
The consequences of natural disasters, such as earthquakes, are evident: death, coordination problems, destruction of infrastructure, and displacement of the population. However, according to empirical research, the impact of a natural disaster on economic activity is mixed. This is relevant for seismic countries such as Chile and New Zealand. This paper contributes to the literature on natural disasters and their economic effects by analyzing the cases of two affected regions within these countries. We employ the synthetic control method to estimate the medium-term impact of two large earthquakes on regional output: the 2010 Maule (Chile) event and the 2010–2011 Canterbury (New Zealand) sequence. We find that Chile and New Zealand experienced opposite economic effects: The GDP per capita of the Canterbury region rose above its counterfactual by about 12
We study the long-term consequences of grade retention in French lower secondary schools by tracking a cohort of students who entered grade 6 in 1995–1996 over a 17-year period. Using propensity score matching as a preprocessing step to balance observable characteristics between retained and non-retained students, we estimate regression-adjusted associations between grade retention and both educational attainment and early labor market outcomes. Our results indicate that grade retention is associated with a substantially lower likelihood of completing secondary school and obtaining a higher education qualification. These negative associations are observed across socioeconomic groups and for both boys and girls. While grade retention does not show a significant association with the probability of being employed, it is associated with significantly lower monthly wages. The wage penalty appears for students from both low and high socioeconomic backgrounds and for both genders. We also find that the timing of grade retention matters for educational outcomes: early retentions, especially in grade 6, are more strongly associated with lower educational attainment, while the wage penalty is observed across all grades of retention.
This paper studies the long-run and transitory dynamics of inflation in Mexico from 2004 to 2023 using 299 disaggregated Consumer Price Index components. A large-dimensional Dynamic Factor Model (DFM) is estimated to extract the common and idiosyncratic components of inflation, characterize their stochastic properties, and evaluate the predictive content of the estimated factors. The results indicate that inflation is largely driven by a common trend arising from a non-stationary common factor, while idiosyncratic components are stationary and reflect short-term fluctuations. Forecast accuracy is assessed under alternative window schemes, factor dimensions, and subsample regimes—including pre-pandemic, pandemic, and post-pandemic periods—and compared with traditional statistical and econometric approaches. The DFM performs particularly well in the post-pandemic phase and over the full sample, systematically outperforming competing models, while remaining competitive during the pandemic despite greater forecast uncertainty. Comparisons with Banco de México’s expert expectations show that factor-based forecasts are often comparable to, and occasionally more accurate than, survey consensus. Overall, the findings highlight the usefulness of large-scale factor models for inflation monitoring and policy analysis.
We investigate the effects of implementation of gasoline standard VI in Chinese cities on air quality using city-level hourly pollutant data for 334 Chinese cities between 2014 and 2019. The results from the difference-in-differences models show that the upgrade from gasoline standard V to VI has significantly improved air quality, with a 15.21
Using approximately 10 million LinkedIn resumes of individuals who graduated from 273 research universities in the USA between 1980 and 2016, this paper explores the spatial dimensions of alumni networks and their associations with first job location choices of graduates. The analysis identifies a positive relationship between the strength of alumni networks and the likelihood of new graduates selecting specific metropolitan areas for their initial employment, after controlling for unobserved destination characteristics, individual attributes, return migration, and origin–destination interactions. In the preferred specification, a one-standard-deviation increase in the log number of alumni at a destination is associated with a 2.256 increase in the relative log odds of choosing that metropolitan area. The associations are particularly strong for alumni connections within the same discipline, gender, or cohort. Additionally, stronger relationships are observed in metropolitan areas characterized by higher information costs and other labor market challenges, as well as for demographic groups that face greater competitive disadvantages. Graduates who are active on professional networks like LinkedIn and those from institutions with established alumni traditions appear to benefit most from these connections. These findings underscore the associations between alumni networks and the geographic employment outcomes of new graduates.
We introduce a new approach to Bayesian inference in potentially non-Gaussian structural vector autoregressions. It relies on the result that the elements of the impact matrix are at least set-identified with narrow bounds under standard assumptions. As a result, an efficient simulation algorithm should be capable of exploring the parameter space, even if only some (or none) of the parameters are identified. We consider very efficient Hamiltonian Monte Carlo (HMC) methods. To exploit potential deviations from Gaussianity, we recommend using a versatile error distribution, which nests a Gaussian distribution as a special case. In this manner, we can infer from the data whether the structural shocks are Gaussian and assess the strength of identification by examining the properties of the estimated shock distributions. Simulations and an empirical application to US fiscal policy demonstrate that non-identification can be easily detected from the marginal posteriors of parameters governing the shapes of the distributions of the structural shocks, even when the data are Gaussian. They also highlight the importance of efficiently accounting for non-Gaussianity.
This paper investigates whether larger colleges generate greater benefits for local innovation. Exploiting college mergers in China and employing a difference-in-differences design, the analysis shows that merged colleges have a greater impact on local innovation than the combined effect of their pre-merger institutions, indicating that larger institutions not only exert stronger innovative influence but also exhibit increasing returns to scale. This rise in local innovation coincides with improvements in the merged colleges’ research productivity, suggesting a potential mechanism linking institutional productivity to regional innovation. Additional analysis reveals that mergers involving only 4-year universities yield the strongest positive effects, indicating that consolidating larger, research-oriented institutions enable more efficient utilization of research resources. Overall, these findings suggest that policymakers seeking to strengthen local innovation capacity and research productivity should consider promoting resource-sharing initiatives or strategic mergers, particularly among 4-year universities, to unlock the full potential of higher education institutions in advancing research excellence and regional economic development.
We examine the critical impact of oil price uncertainty on corporate innovation culture. We employ a unique measure of corporate innovation derived from textual analysis of earnings call transcripts, offering a more comprehensive perspective on innovation than traditional metrics like R D expenditures and patents. The findings reveal a significant negative effect of oil price uncertainty on corporate innovation, indicating that firms tend to scale back innovation activities in response to heightened uncertainty. Since we account for overall economic uncertainty, the observed effect of oil price uncertainty on innovation is distinct and goes beyond the influence of general economic conditions. Importantly, firms where innovation is more critical to profitability experience a smaller decline in innovation when faced with oil price uncertainty. This research contributes to the understanding of how external economic factors, particularly oil price volatility, influence corporate innovation strategies and highlights the differential impact across firms based on the strategic importance of innovation.
Does stronger employment protection reduce income inequality? Answering this question is empirically challenging because labor market institutions and inequality are both highly persistent over time and likely jointly determined. This paper addresses these challenges by developing an econometric framework for recovering long-run institutional relationships, combining a two-stage dynamic panel approach with a copula-based correction for endogenous regressors. We apply this framework to a dynamic panel of 32 OECD countries over the period 1985–2019. Our results suggest that stronger employment protection legislation is associated with lower income inequality, both before and after redistribution, with particularly robust equalizing effects stemming from stronger protection of regular contracts. Beyond the EPL application, the paper contributes methodologically by offering an empirical strategy for estimating the effects of persistent institutional variables in dynamic panel settings.
Maintaining investment efficiency has become increasingly important for firms in an environment characterized by elevated uncertainty. This study examines the relationship between firm-level uncertainty and investment inefficiency of Chinese listed firms, highlighting the crucial role of corporate internal control in preserving investment efficiency. Using newly constructed measures of firm-level uncertainty and an unbalanced quarterly firm-level panel data covering the period from 2009 to 2022, the results show that heightened firm-level uncertainty worsens investment inefficiency. However, the adverse effect is mitigated for firms with a robust internal control mechanism. Mechanism analysis shows that internal control offsets the effect of uncertainty on investment inefficiency by reducing corporate underinvestment. Moreover, analysis using disaggregated internal control indicators suggests that strategic planning, operating management, report reliability, legal compliance, and asset safety play a significant role in alleviating the effect of uncertainty on investment inefficiency . Heterogeneity analysis reveals that the moderating effect of internal control is significant for state-owned enterprises (SOEs), but insignificant for foreign-funded firms. Overall, this study highlights the importance of strengthening corporate internal control in preserving investment efficiency during periods of heightened uncertainty.
I analyze the use of group variation in the probability of receiving a treatment to recover information about the causal effects of that treatment, within the framework of a structural model of counterfactual outcomes and enrollment into the treatment. Specifically, I study the conditions under which a simple group difference in differences—that is, subtracting the difference in mean outcomes between treated and untreated units belonging to a group with a low treatment rate from that same difference for units belonging to a group with a high rate—identifies a lower bound on the difference in average treatment effects between the high- and low-rate groups. While the group difference in treatment effects is directly informative about inequality and heterogeneity in treatment effects, when theory or prior empirical evidence implies that the effect of the treatment is nonnegative, this group difference in differences also identifies a lower bound on the average treatment effect itself for the high-rate group. Although the conditions required for the lower-bound argument are not directly verifiable, I suggest falsification tests for whether they are consistent with the data. I also present several examples, illustrating the applicability of the identification results and the effectiveness of the falsification tests.
In the context of a structural GARCH-in-Mean VAR model, we investigate the effects of monetary uncertainty on unemployment rates by race and ethnicity in the United States. We find that monetary uncertainty tends to increase unemployment across all levels, with the magnitude of the effect being much larger for Black and Hispanic workers than for their White counterparts. Using impulse response analysis, we also provide evidence that a negative real money supply shock tends to dampen employment across all levels (race and different age cohorts based on gender). We find that in the prime working age group of 25–54 years, the female unemployment rate typically reacts more strongly to shocks involving uncertainty in the real money supply than do male unemployment rates. We also find that for both males and females, the youngest and inexperienced workers are the ones who suffer the most from volatility in the real money supply. We find that monetary uncertainty not only increases unemployment but also exacerbates age, gender, and racial employment disparities.
Policy measures restricting international and domestic travel during the COVID-19 pandemic led to substantial distortions in tourism demand data, which continue to affect forecasting models trained on such samples. To address this issue, we employ a random forest approach to forecast monthly tourism demand using lagged values of the target variable and ex ante observable calendar variables as predictors. This specification allows the model to exploit both historic observations of tourism demand and forward-looking information. In an application to Austrian tourism demand, we find that random forests perform competitively prior to COVID-19 and clearly outperform benchmark methods after the pandemic. This improvement can be attributed to the ability of random forests to capture nonlinear relationships between tourism demand and its predictors, as well as to a shift in predictive relevance from historic observations toward calendar variables in the aftermath of the pandemic. Beyond the tourism forecasting application, the results suggest that random forests may also be useful for forecasting seasonal time series more generally.
Extensive research has focused on machine learning applications in credit decision making; yet, systematic quantification of the trade-off between algorithmic predictive accuracy and fairness remains lacking. This paper leverages variations in fairness constraint intensity to identify the efficiency costs of algorithmic fairness in credit markets. Based on 328,600 loan records from Lending Club, this study systematically adjusted fairness constraints across three algorithms—logistic regression, random forest, and XGBoost—to construct complete Pareto frontiers. Through SHAP value decomposition, this paper reveals how seemingly neutral features generate systematic discrimination. The results present two key findings: First, this study documents a positive correlation between algorithmic complexity and discrimination levels consistent with statistical discrimination theory—XGBoost improves AUC by 11.4
We employ a regression discontinuity design (RDD) to examine the effect of halving income tax on the quality of export products among small and microenterprises, leveraging the preferential tax policy introduced in 2014 as a quasi-natural experiment. A 50
We construct a novel dataset on Apple product prices to assess the adherence of different price measures to the law of one price (LOP) and purchasing power parity (PPP), along with their convergence properties following deviations. Using panel methods, with Bayesian estimates retained as supplementary robustness evidence, we evaluate overall conformity to LOP/PPP and then apply dynamic models to measure price and exchange rate convergence. We estimate the half-lives of deviations and quantify the adjustment process by examining how the adjustment parameters evolve over time. Our findings show that Apple products adhere more closely to LOP/PPP than traditional price measures, such as the Big Mac Index and the Consumer Price Index, with deviations lasting only a few months. Transaction costs are identified as a primary driver of persistent deviations from LOP, providing new empirical support for its theoretical predictions. These results highlight the advantages of using homogeneous, globally traded goods like Apple products in studying real exchange rate dynamics.
