Sunday, September 24, 2017

Egalitarian LASSO for Forecast Combination

Here's a new one.  It was something of a long and winding road.  We introduce simple "egalitarian LASSO" procedures that set some combining weights to zero and shrink those remaining toward equality.  The feasible versions don't work very well, due do difficulties associated with cross-validating tuning parameters in small samples, but the lessons learned in studying the infeasible version turn out to be very valuable -- indeed they directly motivate a new procedure, which we call "best <N-averaging", which solves the cross-validation problem and performs intriguingly well.

Diebold, F.X. and Shin, M. (2017), “Beating the Simple Average:  Egalitarian LASSO for Combining Economic Forecasts”, Penn Institute for Economic Research (PIER) Working Paper No. 17-017, available at SSRN:

Friday, September 22, 2017

National Bank of Poland

It strikes me that I'm seeing progressively more research in dynamic predictive modeling from the National Bank of Poland.  A few recent examples appear below.  Related information is here.  Nice job.

Karol Szafranek
Bagged artificial neural networks in forecasting inflation: An extensive comparison with current modelling frameworks

Siem Jan Koopman
André Lucas
Marcin Zamojski

Dynamic term structure models with score-driven time-varying parameters: estimation and forecasting

Piotr Bańbuła
Marcin Pietrzak

Early warning models of banking crises applicable to non-crisis countries

Alessia Paccagnini
Forecasting with FAVAR: macroeconomic versus financial factors

Paweł Pońsko
Bartosz Rybaczyk

Fan chart – a tool for NBP’s monetary policy making

Halina Kowalczyk
Ewa Stanisławska

Are experts’ probabilistic forecasts similar to the NBP projections?

Sunday, September 17, 2017

Machine Learning Meets Central Banking

Here's a nice new working paper from the Bank of England.  There's nothing new methodologically, but there are three fascinating and detailed applications / case studies (banking supervision under imperfect information, UK CPI inflation forecasting, unicorns in financial technology).  For your visual enjoyment I include their Figure 19 below.  (It's the network graph for global technology start-ups in 2014, not spin-art...)

Monday, September 11, 2017

2017 NBER-NSF Time Series Meeting

Just back from 2017 NBER-NSF Time Series at Northwestern.  Quite a feast -- my head is spinning.  Program dumped below; formatted version here.  Many thanks to the program committee for producing this event, and more generally for keeping the series going, year after year, stronger than ever.  (See here for some history and links to past locations, programs, etc.)

The papers were very strong.  Among those that I found particularly interesting are:

-- Moon.  Forecasting in short panels.  You'd think it would be impossible since you need the individual effects.  But it's not.

“Forecasting with Dynamic Panel Data Models”, Hyungsik Roger Moon (University of Southern California), Laura Liu, and Frank Schorfheide

-- Shephard.  Causal estimation meets time series.

“Time series experiments, causal estimands and exact p-values”, Neil Shephard (Harvard University) and Iavor Bojinov

-- The entire (and marvelously-coherent) "Lumsdaine Sesssion" (Pruitt, Pelger, Giglio).  Real progress on econometric methods for identifying financial-market risk factors, with sharp empirical results. 

“Instrumented Principal Component Analysis”, Seth Pruitt (Arizona State University), Bryan Kelly, and Yinan Su
“Estimating Latent Asset-Pricing Factors”, Markus Pelger (Stanford University) and Martin Lettau

“Inference on Risk Premia in the Presence of Omitted Factors”, Stefano Giglio (University of Chicago) and Dacheng Xiu


2017 NBER-NSF Time Series Conference
Friday, September 8 – Saturday, September 9
Kellogg School of Management
Kellogg Global Hub
2211 N Campus Drive; Evanston, IL 60208
Friday, September 8
Registration begins 10:20am (White Auditorium)
Welcome and opening remarks: 10:50am
Session 1: 11:00am – 12:30pm
Chair: Ruey S. Tsay (University of Chicago)
 “Egalitarian Lasso for Shrinkage and Selection in Forecast Combination” Francis X. Diebold (University of Pennsylvania) and Minchul Shin
 “Forecasting with Dynamic Panel Data Models” Hyungsik Roger Moon (University of Southern California), Laura Liu, and Frank Schorfheide
 “Large Vector Autoregressions with Stochastic Volatility and Flexible Priors” Andrea Carriero (Queen Mary University of London), Todd E. Clark, and Massimiliano Marcellino
12:30pm - 2:00pm: Lunch and Poster Session 1 (Faculty Summit, 4th Floor)
 “The Dynamics of Expected Returns: Evidence from Multi-Scale Time Series Modeling“ Daniele Bianchi (University of Warwick)
 “Testing for Unit-root Non-stationarity against Threshold Stationarity” Kung-Sik Chan (University of Iowa)
 “Group Orthogonal Greedy Algorithm for Change-point Estimation of Multivariate Time Series” Ngai Hang Chan (The Chinese University of Hong Kong)
 “The Impact of Waiting Times on Volatility Filtering and Dynamic Portfolio Allocation” Dobrislav Dobrev (Federal Reserve Board of Governors)
 “Testing for Mutually Exciting Jumps and Financial Flights in High Frequency Data” Mardi Dungey (University of Tasmania), Xiye Yang (Rutgers University) presenting
 “Pockets of Predictability” Leland E. Farmer (University of California, San Diego)
 “Factor Models of Arbitrary Strength” Simon Freyaldenhoven (Brown University)
 “Inference for VARs Identified with Sign Restrictions” Eleonora Granziera (Bank of Finland)
 “The Time-Varying Effects of Conventional and Unconventional Monetary Policy: Results from a New Identification Procedure” Atsushi Inoue (Vanderbilt University)
 “On spectral density estimation via nonlinear wavelet methods for non-Gaussian linear processes” Linyuan Li (University of New Hampshire)
 “Multivariate Bayesian Predictive Synthesis in Macroeconomic Forecasting” Kenichiro McAlinn (Duke University)
 “Periodic dynamic factor models: Estimation approaches and applications” Vladas Pipiras (University of North Carolina)
 “Canonical stochastic cycles and band-pass filters for multivariate time series” Thomas M. Trimbur (U. S. Census Bureau)
Session 2: 2:00pm - 3:30pm
Chair: Giorgio Primiceri (Northwestern University)
 “Understanding the Sources of Macroeconomic Uncertainty” Tatevik Sekhposyan (Texas A&M University), Barbara Rossi, and Matthieu Soupre
 “Safety, Liquidity, and the Natural Rate of Interest” Marco Del Negro (Federal Reserve Bank of New York), Domenico Giannone, Marc P. Giannoni, and Andrea Tambalotti
 “Structural Interpretation of Vector Autoregressions with Incomplete Identification: Revisiting the Role of Oil Supply and Demand Shocks” Christiane Baumeister (University of Notre Dame) and James D. Hamilton
Afternoon Break: 3:30pm-4:00pm
Session 3: 4:00pm – 5:30pm
Chair: Serena Ng (Columbia University)
 “Controlling the Size of Autocorrelation Robust Tests” Benedikt M. Pötscher (University of Vienna) and David Preinerstorfer
 “Heteroskedasticity Autocorrelation Robust Inference in Time Series” Regressions with Missing Data Timothy J. Vogelsang (Michigan State University) and Seung-Hwa Rho
 “Time series experiments, causal estimands and exact p-values” Neil Shephard (Harvard University) and Iavor Bojinov
5:30pm – 7pm: Cocktail Reception and Poster Session 2 (Faculty Summit, 4th Floor)
 “Macro Risks and the Term Structure of Interest Rates” Andrey Ermolov (Fordham University)
 “Holdings-based Fund Performance Measures: Estimation and Inference” Wayne E. Ferson (University of Southern California), Junbo L. Wang (Louisiana State University) presenting
 “Economic Predictions with Big Data: The Illusion of Sparsity” Domenico Giannone (Federal Reserve Bank of New York)
 “Estimation and Inference of Dynamic Structural Factor Models with Over-identifying Restrictions” Xu Han (City University of Hong Kong)
 “Bayesian Predictive Synthesis: Forecast Calibration and Combination” Matthew C. Johnson (Duke University)
 “Time Series Modeling on Dynamic Networks” Jonas Krampe (TU Braunschweig)
 “The Complexity of Bank Holding Companies: A Topological Approach” Robin L. Lumsdaine (American University)
 “Sieve Estimation of Option Implied State Price Density” Zhongjun Qu (Boston University) - Junwen Lu (Boston University) presenting
 “Linear Factor Models and the Estimation of Expected Returns” Cisil Sarisoy (Northwestern University)
 “Efficient Parameter Estimation for Multivariate Jump-Diffusions” Gustavo Schwenkler (Boston University)
 “News-Driven Uncertainty Fluctuations” Dongho Song (Boston College)
 “Contagion, Systemic Risk and Diagnostic Tests in Large Mixed Panels” Cindy S.H. Wang (National Tsing Hua University and CORE, University Catholique de Louvain)
7-10pm: Dinner (White Auditorium)
 Dinner speaker: Nobel Laureate Robert F. Engle
Saturday, September 9
Continental Breakfast: 8:00am – 8:30am
Registration begins 8:30am (White Auditorium)
Session 4: 9:00am – 10:30am
Chair: Thomas Severini (Northwestern University)
 “Estimation of time varying covariance matrices for large datasets” Liudas Giraitis (Queen Mary University of London), Y. Dendramis, and G. Kapetanios
 “Indirect Inference With(Out) Constraints” Eric Renault (Brown University) and David T. Frazier
 “Edgeworth expansions for a class of spectral density estimators and their applications to interval estimation” S.N. Lahiri (North Carolina State University) and A. Chatterjee
Morning Break: 10:30am-11:00am
Session 5: 11:00am-12:30pm
Chair: Robin L. Lumsdaine (American University)
 “Instrumented Principal Component Analysis” Seth Pruitt (Arizona State University), Bryan Kelly, and Yinan Su
 “Estimating Latent Asset-Pricing Factors” Markus Pelger (Stanford University) and Martin Lettau
 “Inference on Risk Premia in the Presence of Omitted Factors” Stefano Giglio (University of Chicago) and Dacheng Xiu
12:30pm-2pm: Lunch and Poster Session 3 (Faculty Summit, 4th Floor)
 “Regularizing Bayesian Predictive Regressions” Guanhao Feng (City University of Hong Kong)
 “Good Jumps, Bad Jumps, and Conditional Equity Premium” Hui Guo (University of Cincinnati)
 “High-dimensional Linear Regression for Dependent Observations with Application to Nowcasting” Yuefeng Han (The University of Chicago)
 “Maximum Likelihood Estimation for Integer-valued Asymmetric GARCH (INAGARCH) Models” Xiaofei Hu (BMO Harris Bank, N.A.)
 “Tail Risk in Momentum Strategy Returns” Soohun Kim (Georgia Institute of Technology)
 “The Perils of Counterfactual Analysis with Integrated Processes” Marcelo C. Medeiros (Pontifical Catholic University of Rio de Janeiro) and Ricardo Masini (Pontifical Catholic University of Rio de Janeiro)
 “Anxious unit root processes” Jon Michel (The Ohio State University)
 “Limiting Local Powers and Power Envelopes of Panel AR and MA Unit Root Tests” Katsuto Tanaka (Gakushuin University)
 “High-Frequency Cross-Market Trading: Model Free Measurement and Applications”
Ernst Schaumburg (AQR Capital Management, LLC) – Dobrislav Dobrev (Federal Reserve Board of Governors) presenting
 “A persistence-based Wold-type decomposition for stationary time series” Claudio Tebaldi (Bocconi University)
 “Necessary and Sufficient Conditions for Solving Multivariate Linear Rational Expectations Models and Factoring Matrix Polynomials” Peter A. Zadrozny (Bureau of Labor Statistics)
Session 6: 2:00pm – 3:30pm
Chair: Beth Andrews (Northwestern University)
 “Models for Time Series of Counts with Shape Constraints” Richard A. Davis (Columbia University) and Jing Zhang
 “Computationally Efficient Distribution Theory for Bayesian Inference of High-Dimensional Dependent Count-Valued Data” Scott H. Holan (University of Missouri, U.S. Census Bureau), Jonathan R. Bradley, and Christopher K. Wikle
 “Functional Autoregression for Sparsely Sampled Data”
Daniel R. Kowal (Cornell University, Rice University)

Monday, September 4, 2017

More on New p-Value Thresholds

I recently blogged on a new proposal heavily backed by elite statisticians to "redefine statistical significance", forthcoming in the elite journal Nature Human Behavior. (A link to the proposal appears at the end of this post.) 

I have a bit more to say. It's not just that I find the proposal counterproductive; I have to admit that I also find it annoying, bordering on offensive.

I find it inconceivable that the authors' p<.005 recommendation will affect their own behavior, or that of others like them. They're all skilled statisticians, hardly so naive as to declare a "discovery" simply because a p-value does or doesn't cross a magic threshold, whether .05 or .005. Serious evaluations and interpretations of statistical analyses by serious statisticians are much more nuanced and rich -- witness the extended and often-heated discussion in any good applied statistics seminar.

If the p<.005 threshold won't change the behavior of skilled statisticians like the proposal's authors, then whose behavior MIGHT it change? That is, reading between the lines, to whom is the proposal REALLY addressed?  Evidently those much less skilled, the proverbial "practitioners", who the authors evidently hope might be kept from trouble by a rule of thumb that can at least be followed mechanically.

How patronizing.


Redefine Statistical Significance

Date: 2017
Daniel Benjamin ; James Berger ; Magnus Johannesson ; Brian Nosek ; E. Wagenmakers ; Richard Berk ; Kenneth Bollen ; Bjorn Brembs ; Lawrence Brown ; Colin Camerer ; David Cesarini ; Christopher Chambers ; Merlise Clyde ; Thomas Cook ; Paul De Boeck ; Zoltan Dienes ; Anna Dreber ; Kenny Easwaran ; Charles Efferson ; Ernst Fehr ; Fiona Fidler ; Andy Field ; Malcom Forster ; Edward George ; Tarun Ramadorai ; Richard Gonzalez ; Steven Goodman ; Edwin Green ; Donald Green ; Anthony Greenwald ; Jarrod Hadfield ; Larry Hedges ; Leonhard Held ; Teck Hau Ho ; Herbert Hoijtink ; James Jones ; Daniel Hruschka ; Kosuke Imai ; Guido Imbens ; John Ioannidis ; Minjeong Jeon ; Michael Kirchler ; David Laibson ; John List ; Roderick Little ; Arthur Lupia ; Edouard Machery ; Scott Maxwell; Michael McCarthy ; Don Moore ; Stephen Morgan ; Marcus Munafo ; Shinichi Nakagawa ; Brendan Nyhan ; Timothy Parker ; Luis Pericchi; Marco Perugini ; Jeff Rouder ; Judith Rousseau ; Victoria Savalei ; Felix Schonbrodt ; Thomas Sellke ; Betsy Sinclair ; Dustin Tingley; Trisha Zandt ; Simine Vazire ; Duncan Watts; Christopher Winship ; Robert Wolpert ; Yu Xie; Cristobal Young ; Jonathan Zinman ; Valen Johnson

Abstract: We propose to change the default P-value threshold for statistical significance for claims of new discoveries from 0.05 to 0.005.