We evaluate the robustness of the regional Kuznets curve using the Bayesian average of classical estimates for panel data and identify the robust determinants of regional inequality. Our simulation exercise suggests that this method recovers the variables underlying the true data generating process. Our results indicate that in addition to real GDP per capita, linear and quadratic, the most robust determinants of regional inequality are natural resource rents, arable land and ethnic inequality. We find an inverted-U-shaped relationship between regional inequality and national development in the range of USD 189 to USD 71,682. Beyond this threshold, there is evidence suggesting inequality stabilization.

The place of residence of unarrested criminals is mostly unknown. Existing research has not yet exploited that arrested criminals are a lower bound for criminals to enhance law enforcement and design structural policies. Based upon the stochastic frontier analysis, we propose a model to identify neighborhoods where unarrested criminals are likelier to live. We illustrate our approach empirically by considering Medellin, Colombia, a natural experimental field to analyze crime. We identify that unarrested murderers and drug dealers often reside in overlapping or neighboring areas with shared risk factors, reflecting the city's history of drug-related violence. In addition, we find that employment policies targeting the young and unemployed living in the central-east and the north can mitigate homicides and motorcycle thefts. These findings illustrate how our proposal can be implemented to strengthen state capacities and design targeted, place-based policies for preventing and mitigating crime.

The response of illicit drug consumers to policy changes like legalization is mediated by demand behavior. Since individual drug use is driven by many unobservable factors, accounting for unobserved heterogeneity becomes crucial for designing targeted policies. This paper introduces a finite Gaussian mixture of EASI demand systems to estimate joint demand for marijuana, cocaine, and basuco (a low-purity cocaine paste) in Colombia, accounting for corner solutions and endogenous prices. Our method classifies users into two groups with distinct preferences over consumption: "soft" and "hard" users. Nationally representative survey estimates find drugs are unit-elastic, with marijuana and cocaine complementary. International marijuana legalization episodes along with Colombia's low marijuana production cost suggest legalization is likely to drive prices down significantly. Legalization counterfactuals under the most likely scenario of a 50% marijuana price decrease reveal $363/year welfare gains for consumers, $120M in governement revenue, and $127M dealer losses.

We use approximate Bayesian computation (ABC) to estimate panel data stochastic frontier models, allowing for persistent and transient inefficiency, unobserved heterogeneity, and noise. We use ABC to estimate the generalized true random-effects (GTRE) specification. Simulation exercises for estimating technical efficiency show that our proposal has good finite-sample properties under different configurations of the variance parameters of the four random components, as well as on five well-known datasets. Our proposal is easy to implement in the half-normal case, and adaptable to different distributional assumptions regarding the one-sided error components.

Increases in the use of Bayesian inference in applied analysis, the complexity of estimated models, and the popularity of efficient Markov chain Monte Carlo (MCMC) inference under conjugate priors have led to more scrutiny regarding the specification of the parameters in prior distributions. Impact of prior parameter assumptions on posterior statistics is commonly investigated in terms of local or pointwise assessments, in the form of derivatives or more often multiple evaluations under a set of alternative prior parameter specifications. This paper expands upon these localized strategies and introduces a new approach based on the graph of posterior statistics over prior parameter regions (sensitivity manifolds) that offers additional measures and graphical assessments of prior parameter dependence. Estimation is based on multiple point evaluations with Gaussian processes, with efficient selection of evaluation points via active learning, and is further complemented with derivative information. The application introduces a strategy to assess prior parameter dependence in a multivariate demand model with a high dimensional prior parameter space, where complex prior-posterior dependence arises from model parameter constraints. The new measures uncover a considerable prior dependence beyond parameters suggested by theory, and reveal novel interactions between the prior parameters and the elasticities.

We perform a welfare analysis due to a tax on electricity consumption based on the incomplete exact affine Stone index (EASI) model using a novel data set in the Colombian economy. We provide a novel inferential framework based on a non-parametric specification of the stochastic errors using Dirichlet processes mixtures that allows handling non-normal errors, gaining efficiency, and taking into account, microeconomic restrictions, censoring, simultaneous endogeneity and non-linearity. We find that there is a 95% probability that the equivalent variation of the representative household is between US¢34.1 and US¢34.3, given an approximately 0.8% tariff increase (US¢0.12 per kWh). In addition, we observe that the welfare loss of the representative household of the lowest socioeconomic characteristics is approximately twice the loss of the representative household of the highest socioeconomic characteristics.

We present a procedure to diagnose model misspecification in situations where inference is performed using approximate Bayesian computation. We demonstrate theoretically, and empirically that this procedure can consistently detect the presence of model misspecification. Our examples demonstrates that this approach delivers good finite-sample performance and is computational less onerous than existing approaches, all of which require re-running the inference algorithm. An empirical application to modelling exchange rate log returns using a g-and-k distribution completes the paper.

It seems that facilitating access to a higher spectrum of schools implies that students will attend higher quality schools, as measured by students' end-of-class test scores. We test this hypothesis showing new evidence for the effects of school transport subsidies targeting low-income students on school choice in the context of a devel-oping country (Colombia) using a unique panel dataset involving a public-school population with approximately 15 million records. We built a creative instrument deducing unobserved optimal commute decisions, which seems to satisfy the exclusion and relevance conditions, and we found by means of two-stage least squares that metro and bus subsidy beneficiaries choose statistically and economically significantly better schools, approxi-mately a 33% and 37% improvement in the quality school index, respectively. In addition, we found using endogenous ordered probit models that these subsidies increase the probability of attending very high-quality schools by 59% and 94% for the representative beneficiary, respectively. These results suggest that the reduc-tion of costs of transport not only increases accessibility and the set of school choices among low-income stu-dents, but also targets students enrolled in better quality schools. Therefore, the local government should increase efforts to get more subsidies targeting uncovered areas.

Despite colossal economic and human losses caused by conflict and violence, designing effective policies to avoid conflict remains challenging. While the literature has proposed a voluminous set of candidate predictors, their robustness is questionable and model uncertainty masks the true drivers of conflicts and wars. Considering a comprehensive set of 34 potential determinants in 175 post-Cold-War countries, we employ stochastic search variable selection (SSVS) to sort through all 234 possible models to address model uncertainty. We find past conflict constitutes the most powerful predictor of current conflict: Path dependency matters. Also, larger shares of Jewish, Muslim, or Christian citizens are associated with increased conflict, while economic and political factors remain less relevant than colonial origin and religion. Our results help future researchers and policymakers by inching towards causality and providing a standard set of covariates that need to be accounted for in designing any relevant policies.

We estimate the X-factor of the Colombian electric power distribution sector in the period 2010-2019 by means of stochastic frontier analysis. Our estimates suggest an overall average X-factor equal to -1.6%, where just 3 out of 23 network operators (NOs) have average positive X-factors. In addition, we found that the average efficiency in the period is equal to 32.8% in this sector. This suggests that there is gap for improvements in the sector.

We examine the effect of an integrity pilot campaign on undergraduates' behavior. As with many costly small-scale experiments and pilot programs, our statistical inference has to rely on small sample size. To tackle this issue, we perform a Bayesian retrospective power analysis. In our setup, a lecturer intentionally makes mistakes that favors students' grades, who decide whether to disclose them or not. We find evidence that at least in the short term, the pilot campaign has a positive impact on the students' disclosure probability.

Proper scoring rules are used to assess the out-of-sample accuracy of probabilistic forecasts, with different scoring rules rewarding distinct aspects of forecast performance. Herein, we re-investigate the practice of using proper scoring rules to produce probabilistic forecasts that are `optimal' according to a given score, and assess when their out-of-sample accuracy is superior to alternative forecasts, according to that score. Particular attention is paid to relative predictive performance under misspecification of the predictive model. Using numerical illustrations, we document several novel findings within this paradigm that highlight the important interplay between the true data generating process, the assumed predictive model and the scoring rule. Notably, we show that only when a predictive model is sufficiently compatible with the true process to allow a particular score criterion to reward what it is designed to reward, will this approach to forecasting reap benefits. Subject to this compatibility however, the superiority of the optimal forecast will be greater, the greater is the degree of misspecification. We explore these issues under a range of different scenarios, and using both artificially simulated and empirical data.

SummaryThis paper proposes a Bayesian approach to perform inference in the exact affine Stone index (EASI) demand system that was proposed by Lewbel and Pendakur (2009), while taking into account nonlinearity and endogeneity. A Bayesian approach enables us to easily handle censored data, test and impose inequality restrictions (strict cost monotonicity) and concavity of the cost function, and perform inference of nonlinear functions of the parameter estimates as by‐product of the posterior chains. We compare our proposal with Lewbel and Pendakur (2009)'s results, based on iterative linear three‐stage least squares (3SLS). Although we found no statistically significant differences in point estimates between these two approaches, it seems that ignoring censoring overestimates precision.

We propose a Bayesian one-stage approach to estimate the effect of inefficiency on the time to failure (bankruptcy) of U.S. commercial banks. We do so combining stochastic frontier and proportional hazards settings. Most of the existing literature use two-stage methods which may yield inefficient, biased, and inconsistent estimates. Our proposal overcomes these issues, allows computing the marginal distribution of inefficiencies for each observational unit, and facilitates statistical inference of non-linear functions of parameters such as returns to scale. Simulation exercises show that our proposal outperforms the two-stage maximum likelihood approach traditionally used in the literature. In addition, empirical evidence suggests that inefficiency of U.S. commercial banks during the global financial crisis in 2008–2009 played a statistically and economically significant role determining the time to failure.