"Hemisphere" Neural Networks

A creative new paper by Phillipe Goulet Coulombe at Université du Québec à Montréal, "A Neural Phillips Curve and a Deep Output Gap," introduces "hemisphere" NNs.  Just look at his Figure 1 and you'll understand the structure instantly.  Of course hemisphere structure is a significant restriction, not without costs, but it may be reasonable in many cases, and then it it delivers significant benefits (not least, easily-interpreted results).  Phil's Phillips curve analysis is a fine example.

Maybe one could allow for long memory in recurrent hemisphere NNs, or "long short-term memory," as mentioned in some recent blogs.  Interesting that both Phil's paper and the Paranthos paper focus on inflation.  Long memory (fractional integration) in inflation was always a classic empirical finding supporting the idea of fractional integration/differencing (e.g., first differencing the log price level to convert to inflation seems not enough, but twice differencing seems too much). 

Forecasting Epidemics

[I am always grateful to Andrew Harvey for keeping me updated on his latest work. The post below is adapted from private correspondence with Andrew, but any errors of commission or omission are mine alone.]

I blogged previously on early-vintage Harvey-Kattuman work. Their latest is "A Farewell to R: Time-Series Models for Tracking and Forecasting Epidemics". (Also see here.) You can think of it as a farewell to R, as their title indicates, or as a vastly-improved approach to tracking R in real time, producing both nowcast point estimates and confidence intervals.

Looks extremely promising, in contrast to traditional/current "best-practice" methods, as documented for example in Spectator magazine on Sage vs reality, as of Saturday (22 January):

        Sage ‘scenarios’ vs actual: an update | The Spectator.

(Sage is a body of distinguished epidemiologists who have been advising the UK government. They have a track record of pessimistic scenarios based on traditional/current "best practice" methods, and they have been spectacularly wrong. At the beginning of last September they were forecasting between 1500 and 7000 UK hospital admissions per day by the end of September. The Harvey-Kattuman model was predicting 1000. In the end it was around 800.)

Intrinsic Time in Finance

Intrinsic Time in Finance

Call for papers:

Abbey Hegne, Allensbach,
Lake Constance, Germany

6 – 7 May 2022

https://www.wiwi.uni-konstanz.de/pohlmeier/conferencesworkshops/conference-on-intrinsic-time-in-finance/

The conference focuses on bringing together leading scholars, researchers and practitioners from areas
of econometrics, mathematical statistics and finance, to discuss the recent developments in the field of
using the intrinsic time perspective and the concept of time deformation in high-frequency finance and
to disseminate new research ideas and results with broad social-economic benefits. Both theoretical and
empirical, but also computational contributions are welcome.

Confirmed invited speakers:

• Torben Andersen (Northwestern University, USA)
• Tim Bollerslev (Duke University, USA)
• Laurent Calvet (EDHEC Business School, France)
• Michel Dacorogna (University of Zurich and Prime Re Solutions, Switzerland)
• Rainer Dalhaus (University of Heidelberg, Germany)
• Dobrislav Dobrev (Federal Reserve Bank, USA)
• Yingjie Dong (Singapore Management University, Singapore)
• Masaaki Fukasawa (Osaka University, Japan)
• Christian Gourieroux (CREST, France and University of Toronto, Canada)
• Nikolaus Hautsch (University of Vienna, Austria)
• Yifan Li (University of Manchester, UK)
• Richard Olsen (Lykke, Switzerland)
• Eric Renault (University of Warwick, UK)
• Roberto Renò (University of Verona, Italy)
• Mathieu Rosenbaum (Ecole Polytechnique, France)
• George Tauchen (Duke University, USA)
• Viktor Todorov (Northwestern University, USA)
• Agnieszka Wylomanska (Wroclaw University of Science and Technology, Poland)

Submission Deadline:
Papers should be sent as pdf-file to intrinsic.time@uni-konstanz.de until February 15th, 2022.
The review process will last approximately two weeks after the submission deadline. Notification
will be sent via email.

Organizers:
Roxana Halbleib (University of Freiburg, Germany)
Winfried Pohlmeier (University of Konstanz, Germany)
Sandra Nolte (Lancaster University Management School, UK)
Ingmar Nolte (Lancaster University Management School, UK)

Conference Homepage:
https://www.wiwi.uni-konstanz.de/pohlmeier/conferencesworkshops/conference-on-intrinsic-time-in-finance/

Term Structure Modeling for Policy Evaluation

Check out "A term structure model with useful factors: assessing the impact of ECB’s unconventional policies from 2014 to 2020," by Kramer (ECB), Nyholm (ECB), and Sahakyan (BIS).

The latest in dynamic Nelson-Siegel / arbitrage-free Nelson-Siegel (DNS/AFNS) modeling.  They cleverly use a short rate rather than a long-rate (level) factor, and they include the nominal long-term "natural rate of interest" as a fourth factor.

Important Issues in Empirical Macro

Call for Papers - Deadline 15 February 2022 View this email in your browser

Call for Papers EURO AREA BUSINESS CYCLE NETWORK (EABCN)  and Deutsche Bundesbank Conference on Challenges in Empirical Macroeconomics since 2020 Eltville am Rhein, Germany May 19-20, 2022

Organisers: Malte Knüppel (Deutsche Bundesbank) Elmar Mertens (Deutsche Bundesbank) Barbara Rossi (ICREA-Pompeu Fabra University, Barcelona GSE, CREI and CEPR) We welcome submissions from a broad range of themes related to the consequences of the Covid-19 shock for empirical macroeconomics. Topics of the conference include but are not limited to   •    Rare events and disaster-type shocks •    Measuring economic uncertainty and its effects •    Nowcasting and forecasting in a changing environment •    Inflation dynamics •    Business cycle measurement   Serena Ng (Columbia University) and Frank Schorfheide (University of Pennsylvania) have confirmed their participation as invited speakers. The deadline for submissions is 9am GMT on Tuesday, February 15, 2022. Authors of successful submissions will be notified by end of March. We intend to offer a hybrid event. The physical meeting will take place in the Deutsche Bundesbank Training Centre in Eltville am Rhein. The event is sponsored by the EABCN and Deutsche Bundesbank. Travel and accommodation expenses will be reimbursed for academic participants, potentially subject to a cap. If required due to the evolution of the pandemic, the conference will become an online event. Information on this conference is also available on the websites of EABCN and Deutsche Bundesbank.   How to Apply: Authors who are CEPR affiliated or already have a CEPR profile can upload their submission by: 1) Log in on the CEPR portal online at https://portal.cepr.org/ 2) Go to https://portal.cepr.org/meetings/1487/info  3) If you are a member of the MEF programme area, click on "Change registration details", complete the requested information and click "Submit information". 4)  If you have a CEPR profile, click on "Step 1: Apply" and complete the requested information and click "Register" Authors who are not CEPR affiliated or do not have a CEPR profile can: 1) Create an online profile here 2) Log in on the CEPR portal online at https://portal.cepr.org/ 3) Go to  https://portal.cepr.org/meetings/1487/info  4) Click on "Step 1: Apply" and complete the requested information and click "Register" If you have any difficulty in applying please contact, Lydia Williams, CEPR Events Officer at lwilliams@cepr.org for assistance, with the subject line '1487- EABCN and Deutsche Bundesbank Conference 2022'

Centre for Economic Policy Research (CEPR) The Centre for Economic Policy Research is a network of over 1500 Research Fellows and Affiliates, based primarily in European universities. The Centre coordinates the research activities of its Fellows and Affiliates and communicates the results to the public and private sectors. CEPR is an entrepreneur, developing research initiatives with the producers, consumers and sponsors of research. Established in 1983, CEPR is a European economics research organisation with uniquely wide-ranging scope and activities. The Centre is pluralist and non-partisan, bringing economic research to bear on the analysis of medium- and long-run policy questions. cepr_org CEPR Vox LinkedIn VoxViews Facebook Copyright © 2022 Euro Area Business Cycle Network, All rights reserved. unsubscribe from this list    update subscription preferences

More on "Long Short Term Memory"

Following up on a recent blog of mine, Marcelo Medeiros kindly alerted me to this very nice paper by Livia Paranhos, "Predicting Inflation with Neural Networks".  She is a Ph.D. candidate in Economics at Warwick, working with NNs in macro forecasting.  I'm still wondering whether/how "Long Short Term Memory" is related to conventional statistical long memory in the sense of fractional integration.

Machine Learning for Stock Return Volatility

This new Filipović-Khalilzadeh paper is quite nice.

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3995529

They do both tree methods and neural nets for realized volatility forecasting, using not only the RV history but also various "standard" observed predictors.  Fine.  But the really interesting thing is their implementation of the "long short-term memory model," which wins all their races: 

https://direct.mit.edu/neco/article/9/8/1735/6109/Long-Short-Term-Memory

Hard to tell if it's really capturing long memory in the statistical/econometric sense (a crucial finding of Andersen et al (2003)), and they don't discuss or even mention statistical/econometric long memory.  Perhaps the workings of the "long short-term memory model" are close to those of the "Corsi approximation" to long memory used in Andersen et al. (2007).

"Sequence-Space Methods" (?)

 

From a recent announcement that just arrived:

Day 3 – Friday, June 17, 2022

Topic: “Advanced sequence-space methods”
Morning lecture session from  09:00 AM - 12:00 NOON (CEST)
Afternoon practice session from 01:30PM - 3:30PM (CEST)
 
In the morning, we go over advanced sequence-space methods, including second moments and Bayesian estimation to macro time series data. Time permitting, we discuss how to incorporate deviations from full information rational expectations in sequence-space models.
 
In the afternoon, we further work with the Sequence-Space-Jacobian toolbox, compute steady states, linear impulse responses and transition dynamics. Time permitting, we also compute second moments and the likelihood.

So, what are "sequence-space methods"?  Are they not just time-series methods, at least as used in the paragraphs above?  Every place you see "sequence-space," just change to "time-series"?

Macro Skewness and Conditional Second Moments

Look out for this one.  Sounds very cool:

Macro Skewness and Conditional Second Moments: Evidence and Theories

Ian Dew-Becker, Northwestern University and NBER

Alireza Tahbaz-Salehi, Northwestern University

Andrea Vedolin, Boston University and NBER

The closest I can find online is: 

http://www.dew-becker.org/documents/-TS-V.pdf

Wonderful that they prominently reference the fundamental early insights of Dan Sichel.  Yet I'm sad that they fail to mention the even earlier insights, on which Sichel builds, of the late great Salih Neftci.  Sort of like doing a paper on the cleansing effects of recessions and not citing Shumpeter.  Yes, Shumpeter was a century ago, but still.  And yes, Sichel of course cites Neftci, so Dew-Becker et al. implicitly cite him, but still.  (And Neftci was less than half a century ago!)  The classic that started it all is here.

Climate Econometrics

2022 EGU have extended their abstract submission deadline to 14th January 2022 at 1300 CET.

On behalf of the convenors: Felix Pretis, Sam Heft-Neal, Susana Campos-Martins, David Stainforth

__________________

The European Geoscience Union (EGU) General Assembly is a major international geoscience conference that has broadened its scope into a widening range of related disciplines over the past few years.

The conference is planning to run a hybrid in-person/online format and runs from 3-8 April 2022.

Now in its 5th year at EGU, we would like to advertise the session (description below): Economics and Econometrics of Climate Change: evaluating the drivers, impacts, and policies of climate change.

Please consider submitting your research - details on how to do so are here.

_____________

CL3.2.3: Economics and Econometrics of Climate Change: evaluating the drivers, impacts, and policies of climate change

Understanding the impact of climate change on natural and socio-economic outcomes plays an important role in informing a range of national and international policies, including energy, agriculture, and health. However economic models of (and those designed to include) climate impacts that guide decision makers rely on multiple components, for example projections of future climate change, damage functions, and policy responses, each of which comes with its own modelling challenges and uncertainties.
We invite research using process-based (e.g., Integrated Assessment Models) and empirical models of climate change to investigate future human and natural impacts, together with policy evaluation to explore effective mitigation, technology, and adaptation pathways. Furthermore, we invite research on changes to, and new developments of climate-economic and econometric modelling.

SoFiE 2022 University of Cambridge CFP

 Call for papers ends soon (Jan 31)!

https://www.janeway.econ.cam.ac.uk/event/fourteenth-annual-sofie-conference

Pierre-Olivier Gourinchas next IMF Chief Economist

 https://urldefense.com/v3/__https://www.imf.org/en/News/Articles/2022/01/10/pr2201-imf-managing-director-names-pierre-olivier-gourinchas-economic-counsellor-head-research-dept__;!!IBzWLUs!A1Fm43j8dP-vOaalZT4jLjSsET0tEaYlzh3voULxIyCOh51IZ0HZRSOvzq-dQOQ-1yWJ$

Approximating Bayes in the 21st Century

 Can't wait to read this!

Approximating Bayes in the 21st Century

Gael M. Martin, David T. Frazier and Christian P. Robert

https://www.monash.edu/business/ebs/research/publications/ebs/wp24-2021.pdf

Abstract

The 21st century has seen an enormous growth in the development and use of approximate Bayesian methods. Such methods produce computational solutions to certain ‘intractable’ statistical problems that challenge exact methods like Markov chain Monte Carlo: for instance, models with unavailable likelihoods, high-dimensional models, and models featuring large data sets. These approximate methods are the subject of this review. The aim is to help new researchers in particular – and more generally those interested in adopting a Bayesian approach to empirical work – distinguish between different approximate techniques; understand the sense in which they are approximate; appreciate when and why particular methods are useful; and see the ways in which they can can be combined.