|
on Econometrics |
| By: | De Vos, Ignace (Department of Economics, Lund University); Stauskas, Ovidijus (Department of Economics, Lund University) |
| Abstract: | The Common Correlated Effects (CCE) methodology is now well established for the analysis of factor-augmented panel models. Yet, it is often neglected that the pooled variant is biased unless the cross-section dimension (N) of the dataset dominates the time series length (T). This is problematic for inference with typical macroeconomic datasets where T often equal or larger than N. Given that an analytical correction is also generally infeasible, the issue remains without a solution. In response, we provide in this paper the theoretical foundation for the cross-section, or pairs bootstrap in large N and T panels with T/N finite. We show that the scheme replicates the distribution of the CCE estimators, under both constant and heterogeneous slopes, such that bias can be eliminated and asymptotically correct inference can ensue even when N does not dominate. Monte Carlo experiments illustrate that the asymptotic properties also translate well to finite samples. |
| Keywords: | Panel data; CCE; Bootstrap; Pairs; Factors; Bias Correction |
| JEL: | C12 C23 C33 |
| Date: | 2021–11–19 |
| URL: | http://d.repec.org/n?u=RePEc:hhs:lunewp:2021_016&r= |
| By: | Kaddoura, Yousef (Department of Economics, Lund University); Westerlund, Joakim (Department of Economics, Lund University) |
| Abstract: | In this article, we propose a new estimator of panel data models with interactive fixed effects and multiple structural breaks that is suitable when the number of time periods, T, is fixed and only the number of cross-sectional units, N, is large. This is done by viewing the determination of the breaks as a shrinkage problem, and to estimate both the regression coefficients, and the number of breaks and their locations by applying a version of the Lasso approach. We show that with probability approaching one the approach can correctly determine the number of breaks and the dates of these breaks, and that the estimator of the regime-specific regression coefficients is consistent and asymptotically normal. We also provide Monte Carlo results suggesting that the approach performs very well in small samples, and empirical results suggesting that the coefficients of the deterrence model of crime are not constant as typically assumed but subject to structural change. |
| Keywords: | Panel data; Interactive effects; Common factors; Structural change; Lasso |
| JEL: | C13 C23 C33 K42 |
| Date: | 2021–11–15 |
| URL: | http://d.repec.org/n?u=RePEc:hhs:lunewp:2021_015&r= |
| By: | Taining Wang (Capital University of Economics and Business); Feng Yao (West Virginia University, Department of Economics) |
| Abstract: | We propose a varying coefficient regression model for panel data that controls for both latent heterogeneities in cross-sectional units and unobserved common shocks over time. The model allows different smoothing variables to enter through either a stand-alone function or a coefficient function. Without requiring a normalization of the fixed effects, we propose a two-step estimator. First, we estimate the varying coefficients with the pilot series-based estimators, eliminating fixed effects though differencing. Second, we perform a one-step kernel backfitting to improve the estimation efficiency. We demonstrate through Monte-Carlo simulations that our estimators are computationally efficient and perform well relative to a profile-based kernel estimator. |
| Keywords: | semiparametric model, varying coefficient model, different smoothing variables, two-way fixed effects, series estimation, kernel backfitting |
| JEL: | C14 C15 C22 |
| Date: | 2021–10 |
| URL: | http://d.repec.org/n?u=RePEc:wvu:wpaper:21-01&r= |
| By: | Jiaxin Guo; Haoyu Wei; Xiaoyu Lei; Jing Luo |
| Abstract: | For the differential privacy under the sub-Gamma noise, we derive the asymptotic properties of a class of network models with binary values with a general link function. In this paper, we release the degree sequences of the binary networks under a general noisy mechanism with the discrete Laplace mechanism as a special case. We establish the asymptotic result including both consistency and asymptotically normality of the parameter estimator when the number of parameters goes to infinity in a class of network models. Simulations and a real data example are provided to illustrate asymptotic results. |
| Date: | 2021–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2111.01301&r= |
| By: | Tae-Hwy Lee (Department of Economics, University of California Riverside); He Wang (University of International Business and Economics, Beijing); Zhou Xi (Citigroup); Ru Zhang (JP Morgan Chase) |
| Abstract: | We consider a multiplicative decomposition of the financial returns to improve the density forecasts of financial returns. The multiplicative decomposition is based on the identity that financial return is the product of its absolute value and its sign. Advantages of modeling the two components are discussed. To reduce the effect of the estimation error due to the multiplicative decomposition in estimation of the density forecast model, we impose a moment constraint that the conditional mean forecast is set to match with the sample mean. Imposing such a moment constraint operates a shrinkage and tilts the density forecast of the decomposition model to produce the improved maximum entropy density forecast. An empirical application to forecasting density of the daily stock returns demonstrates the benefits of using the decomposition and imposing the moment constraint to obtain the improved density forecast. We evaluate the density forecast by comparing the logarithmic score, the quantile score, and the continuous ranked probability score. We contribute to the literature on the density forecast and the decomposition models by showing that the density forecast of the decomposition model can be improved by imposing a sensible constraint in the maximum entropy framework. |
| Keywords: | Decomposition, Copula, Moment constraint, Maximum entropy, Density forecast, Logarithmic score, Quantile score, VaR, Continuous ranked probability score. |
| JEL: | C1 C3 C5 |
| Date: | 2021–11 |
| URL: | http://d.repec.org/n?u=RePEc:ucr:wpaper:202115&r= |
| By: | Sepideh Mosaferi; Mark S. Kaiser |
| Abstract: | This article focuses on cointegrating regression models in which covariate processes exhibit long range or semi-long range memory behaviors, and may involve endogeneity in which covariate and response error terms are not independent. We assume semi-long range memory is produced in the covariate process by tempering of random shock coefficients. The fundamental properties of long memory processes are thus retained in the covariate process. We modify a test statistic proposed for the long memory case by Wang and Phillips (2016) to be suitable in the semi-long range memory setting. The limiting distribution is derived for this modified statistic and shown to depend only on the local memory process of standard Brownian motion. Because, unlike the original statistic of Wang and Phillips (2016), the limit distribution is independent of the differencing parameter of fractional Brownian motion, it is pivotal. Through simulation we investigate properties of nonparametric function estimation for semi-long range memory cointegrating models, and consider behavior of both the modified test statistic under semi-long range memory and the original statistic under long range memory. We also provide a brief empirical example. |
| Date: | 2021–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2111.00972&r= |
| By: | Aibo Gong |
| Abstract: | It is often the case in program evaluation that units will often anticipate the implementation of a new policy before it occurs. Such anticipatory behavior can lead to units' outcomes becoming dependent on their future treatment assignments. In this paper, I employ a potential outcomes framework to analyze the treatment effect with anticipation. I start with a classical difference-in-differences model with two time periods and provide partially identified sets with easy-to-implement estimation and inference strategies for causal parameters. I consider generalizations on including covariates and longitudinal models. I also analyze cases with imperfect anticipation and nonlinear outcomes. I further illustrate my results by analyzing the effect of an early retirement incentive program for teachers, which was likely to be anticipated by the target units, on student achievement. The empirical results demonstrate the potential pitfalls of failing to consider anticipation in policy evaluation. |
| Date: | 2021–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2111.06573&r= |
| By: | Narayanan, Sridhar (Stanford U); Kalyanam, Kirthi (Santa Clara U) |
| Abstract: | The availability of behavioral and other data on customers and advances in machine learning methods have enabled targeting of customers in a variety of domains, including pricing, advertising, recommendation systems and personal selling contexts. Typically, such targeting involves first training a machine learning algorithm on a training dataset, and then using that algorithm to score current or potential customers. When the score crosses a threshold, a treatment (such as an offer, an advertisement or a recommendation) is assigned. In this paper, we demonstrate that this has given rise to opportunities for causal measurement of the effects of such targeted treatments using regression discontinuity designs (RDD). Investigating machine learning in a regression discontinuity framework leads to several insights. First, we characterize conditions under which regression discontinuity designs can be used to measure not just local average treatment effects (LATE), but also average treatment effects (ATE). In some situations, we show that RD can be used to find bounds on the ATE even if we are unable to find point estimates. We then apply this to the machine learning based targeting contexts by studying two different ways in which the score required for targeting is generated, and explore the utility of RDD to these contexts. Finally, we apply our approach in the empirical context of the targeting of retargeted display advertising. Using a dataset from a context where a machine learning based targeting policy was employed in parallel with a randomized controlled trial, we examine the performance of the RDD estimate in estimating the treatment effect, validate it using a placebo test and demonstrate its practical utility. |
| Date: | 2020–12 |
| URL: | http://d.repec.org/n?u=RePEc:ecl:stabus:3925&r= |
| By: | Dong Hwan Oh; Andrew J. Patton |
| Abstract: | Many important economic decisions are based on a parametric forecasting model that is known to be good but imperfect. We propose methods to improve out-of-sample forecasts from a mis- speci ed model by estimating its parameters using a form of local M estimation (thereby nesting local OLS and local MLE), drawing on information from a state variable that is correlated with the misspeci cation of the model. We theoretically consider the forecast environments in which our approach is likely to o¤er improvements over standard methods, and we nd signi cant fore- cast improvements from applying the proposed method across distinct empirical analyses including volatility forecasting, risk management, and yield curve forecasting. |
| Keywords: | Model misspecification; Local maximum likelihood; Volatility forecasting; Value-at-risk and expected shortfall forecasting; Yield curve forecasting |
| JEL: | C53 C51 C58 C14 |
| Date: | 2021–11–05 |
| URL: | http://d.repec.org/n?u=RePEc:fip:fedgfe:2021-71&r= |
| By: | Pallavi Basu; Luella Fu; Alessio Saretto; Wenguang Sun |
| Abstract: | In sparse large-scale testing problems where the false discovery proportion (FDP) is highly variable, the false discovery exceedance (FDX) provides a valuable alternative to the widely used false discovery rate (FDR). We develop an empirical Bayes approach to controlling the FDX. We show that for independent hypotheses from a two-group model and dependent hypotheses from a Gaussian model fulfilling the exchangeability condition, an oracle decision rule based on ranking and thresholding the local false discovery rate (lfdr) is optimal in the sense that the power is maximized subject to FDX constraint. We propose a data-driven FDX procedure that emulates the oracle via carefully designed computational shortcuts. We investigate the empirical performance of the proposed method using simulations and illustrate the merits of FDX control through an application for identifying abnormal stock trading strategies. |
| Keywords: | Cautious Data Mining; False Discovery Exceedance Control; Local False Discovery Rates; Multiple Hypotheses Testing; Poisson Binomial Distribution; Trading Strategies |
| JEL: | C11 C12 C15 |
| Date: | 2021–11–18 |
| URL: | http://d.repec.org/n?u=RePEc:fip:feddwp:93384&r= |
| By: | Vito Polito; Yunyi Zhang |
| Abstract: | We develop a regime switching vector autoregression where artificial neural networks drive time variation in the coefficients of the conditional mean of the endogenous variables and the variance covariance matrix of the disturbances. The model is equipped with a stability constraint to ensure non-explosive dynamics. As such, it is employable to account for nonlinearity in macroeconomic dynamics not only during typical business cycles but also in a wide range of extreme events, like deep recessions and strong expansions. The methodology is put to the test using aggregate data for the United States that include the abnormal realizations during the recent Covid-19 pandemic. The model delivers plausible and stable structural inference, and accurate out-of-sample forecasts. This performance compares favourably against a number of alternative methodologies recently proposed to deal with large outliers in macroeconomic data caused by the pandemic. |
| Keywords: | nonlinear time series, regime switching models, extreme events, Covid-19, macroeconomic forecasting |
| JEL: | C45 C50 E37 |
| Date: | 2021 |
| URL: | http://d.repec.org/n?u=RePEc:ces:ceswps:_9395&r= |
| By: | Michael Greenacre |
| Abstract: | Compositional data analysis is concerned with multivariate data that have a constant sum, usually 1 or 100%. These are data often found in biochemistry and geochemistry, but also in the social sciences, when relative values are of interest rather than the raw values. Recent applications are in the area of very high-dimensional "omics" data. Logratios are frequently used for this type of data, i.e. the logarithms of ratios of the components of the data vectors. These ratios raise interesting issues in matrix-vector representation, computation and interpretation, which will be dealt with in this chapter. |
| JEL: | C19 C88 |
| Date: | 2021–11 |
| URL: | http://d.repec.org/n?u=RePEc:upf:upfgen:1805&r= |
| By: | Konrad Menzel |
| Abstract: | We provide asymptotic approximations to the distribution of statistics that are obtained from network data for limiting sequences that let the number of nodes (agents) in the network grow large. Network formation is permitted to be strategic in that agents' incentives for link formation may depend on the ego and alter's positions in that endogenous network. Our framework does not limit the strength of these interaction effects, but assumes that the network is sparse. We show that the model can be approximated by a sampling experiment in which subnetworks are generated independently from a common equilibrium distribution, and any dependence across subnetworks is captured by state variables at the level of the entire network. Under many-player asymptotics, the leading term of the approximation error to the limiting model established in Menzel (2015b) is shown to be Gaussian, with an asymptotic bias and variance that can be estimated consistently from a single network. |
| Date: | 2021–11 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2111.01678&r= |
| By: | Florens, Jean-Pierre; Enache, Andreea; Sbaï, Erwann |
| Date: | 2021–11–16 |
| URL: | http://d.repec.org/n?u=RePEc:tse:wpaper:126172&r= |