Automated Discovery in Econometrics

Abstract

Our subject is the notion of automated discovery in econometrics. Advances in computer power, electronic communication, and data collection processes have all changed the way econometrics is conducted. These advances have helped to elevate the status of empirical research within the economics profession in recent years and they now open up new possibilities for empirical econometric practice. Of particular significance is the ability to build econometric models in an automated way according to an algorithm of decision rules that allow for (what we call here) heteroskedastic and autocorrelation robust (HAR) inference. Computerized search algorithms may be implemented to seek out suitable models, thousands of regressions and model evaluations may be performed in seconds, statistical inference may be automated according to the properties of the data, and policy decisions can be made and adjusted in real time with the arrival of new data. We discuss some aspects and implications of these exciting, emergent trends in econometrics.

Finite-sample corrected inference for two-step GMM in time series

“…This paper is also related to the HAR literature which is pioneered by Kiefer and Vogelsang (2002a, 2002b, Phillips (2005), Müller (2007) and Sun et al (2008). Recent research along this line can be found in Sun (2014 a&b), Müller and Watson (2018) and Lazarus et al (2019).…”

Testing Convergence Using HAR Inference

“…As (n; T ) ! 1 the statistics t 1 ; t 2 ; t HAR and t HAC are all negative under convergence since their signs are determined by the trend regression coe¢ cient^ nT which is always negative when > 0: Except for a few speci…c cases shown in the statement of the theorem, the asymptotic orders of the t-ratios that rely on HAR estimates can be obtained by replacing by unity in (22). This accords with the understanding that the HAR statistics rely on the use of lag truncation parameters proportional to the sample size.…”

Boosting: Why You Can Use the Hp Filter

“…Decomposition into trend and cycle is commonly achieved by regression or filtering. The latter is primarily motivated by the prior view that a trend is distinguished as a smoothly varying component in relation to the observed data, a concept reflected in the header quotation from Hodrick and Prescott (1997) (hereafter HP), leading to the so‐called HP filter, to the use of spectral methods (Hannan, 1963; Christiano and Fitzgerald, 2003), and to the use of orthonormal polynomial regression (Phillips, 1998, 2005, 2014). The HP filter belongs to the statistical approach that was introduced by Whittaker (1923) and Whittaker and Robinson (1924), who provided a probabilistic framework of penalized maximum likelihood estimation to deliver a quantitative measure of trend (or graduation in their terminology).…”

“…For example, in empirical analysis using country‐level macroindicators, macroeconomists handle data collected from a large number of heterogeneous economies in highly varying stages of economic development. In such cases, where there are many possible determining factors and diverse time series trajectories, automated econometric procedures (Phillips, 2005) can be of tremendous appeal in practical work, as has been argued recently by many authors since the development of high‐dimensional regression methods such as Lasso (Tibshirani, 1996).…”