https://ebookmass.com/product/practical-microsimulation-
Instant digital products (PDF, ePub, MOBI) ready for you
Download now and discover formats that fit your needs...
The ADA Practical Guide to Dental Implants 1st Edition Luigi O. Massa
https://ebookmass.com/product/the-ada-practical-guide-to-dentalimplants-1st-edition-luigi-o-massa/
ebookmass.com
Reproducible Econometrics Using R Jeffrey S. Racine
https://ebookmass.com/product/reproducible-econometrics-using-rjeffrey-s-racine/
ebookmass.com
Introduction to Econometrics 3rd Edition James H. Stock
https://ebookmass.com/product/introduction-to-econometrics-3rdedition-james-h-stock/
ebookmass.com
https://ebookmass.com/product/github_copilot_for_java_developerspritesh-mistry/
ebookmass.com
Eat Without Fear: Harnessing Science to Confront and Overcome Your Eating Disorder Nicholas R. Farrell
https://ebookmass.com/product/eat-without-fear-harnessing-science-toconfront-and-overcome-your-eating-disorder-nicholas-r-farrell/
ebookmass.com
Get Funded! The Startup Entrepreneur's Guide to Seriously Successful Fundraising John Biggs
https://ebookmass.com/product/get-funded-the-startup-entrepreneursguide-to-seriously-successful-fundraising-john-biggs/
ebookmass.com
Macroeconomics, 11th Global Edition Andrew Abel
https://ebookmass.com/product/macroeconomics-11th-global-editionandrew-abel/
ebookmass.com
From Freedom Fighters to Jihadists: Human Resources of Non-State Armed Groups Vera Mironova
https://ebookmass.com/product/from-freedom-fighters-to-jihadistshuman-resources-of-non-state-armed-groups-vera-mironova/
ebookmass.com
Psychic Whispers: Psychic Mystery Romance (Woodward Hill Mystery Romance Book 1) Arial Burnz
https://ebookmass.com/product/psychic-whispers-psychic-mysteryromance-woodward-hill-mystery-romance-book-1-arial-burnz/
ebookmass.com
The Wiley Handbook of Healthcare Treatment Engagement: Theory, Research, and Clinical Practice Andrew Hadler (Editor)
https://ebookmass.com/product/the-wiley-handbook-of-healthcaretreatment-engagement-theory-research-and-clinical-practice-andrewhadler-editor/
ebookmass.com
Practical Microsimulation Modelling
Cathal O’Donoghue
Great Clarendon Street, Oxford, OX2 6DP, United Kingdom
Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries
© Cathal O’Donoghue 2021
The moral rights of the author have been asserted
First Edition published in 2021
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above
You must not circulate this work in any other form and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2021935786
ISBN 978–0–19–885287–2
DOI: 10.1093/oso/9780198852872.001.0001
Printed and bound by CPI Group (UK) Ltd, Croydon, CR0 4YY
Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.
Dedicated to my peers and students in the microsimulation community
Preface
This book describes the lessons that I have learnt over the course of developing my skills as a microsimulation modeller and co-generating knowledge and experience with Ph.D. and master’s students and with industry clients.
While I have built many models over the course of my career, the research in this book draws on lessons from four models:
• SWITCH–the tax-benefit microsimulation model of the Economic and Social Research Institute, in Dublin, Ireland (Callan et al. 1996)
• EUROMOD–the European tax-benefit model (Immervoll et al. 1999; Immervoll and O’Donoghue 2009; Sutherland and Figari 2013)
• LIAM–Lifecycle Income Analysis Model (dynamic microsimulation framework) (O’Donoghue et al. 2009; de Menten et al. 2014)
• SMILE–Simulation Model of the Irish Local Economy (spatial microsimulation model) (O’Donoghue et al. 2012)
Figure 0.1 describes my knowledge-tree equivalent to a genealogical tree of these models and their link to earlier models in the literature.
These models span many of the main methodological areas of microsimulation modelling, covering static, dynamic, cross-country, and spatial models. Starting with static modelling, which ignores behaviour, time, and place, the first model I worked on was the SWITCH static microsimulation model at the Economic and Social Research Institute, in Dublin, Ireland, working with Tim Callan and his team to develop and use the Irish SWITCH static tax-benefit model (Callan et al. 1996). This work was heavily influenced by the Institute for Fiscal Studies’ Taxben model (Blundell et al. 2000).
Extending the dimensionality of the modelling to incorporate the time dimension in a dynamic microsimulation model, I developed a dynamiccohort model LIAM (O’Donoghue 2002) during a Ph.D. at the London School of Economics (LSE), under the supervision of Celia Phillips and Jane Falkingham. LIAM built upon the family of dynamic-cohort models developed at the LSE for the UK (LIFEMOD, Falkingham and Hills 1995) and Australia (HARDING, Harding 1993). This model was also influenced by the work of Steve Caldwell at Cornell University (where I had the pleasure of spending some time during the course of my Ph.D.), who developed the CORSIM model (Caldwell 1996). Caldwell’s model is a direct link to the first microsimulation models in the field, when he partnered with Guy Orcutt (Orcutt et al. 1958; Orcutt 1960) on the DYNASIM model at the Urban Institute, Washington DC, in the early 1970s.
Caldwell had a particular way of building very sophisticated models with relatively limited resources, working with Ph.D. students and with partners in other countries. One of these partnerships was with the Canadian government to build the DYNACAN model, along with Rick Morrison and his team (Caldwell and Morrison 2000). The joint meetings between CORSIM–DYNACAN teams were one of the leading fora for the exchange of knowledge in dynamic microsimulation modelling at the time.
I also worked with Holly Sutherland’s team at the Microsimulation Unit, in the Department of Applied Economics, University of Cambridge, together with partners such as Tony Atkinson and Francois Bourguignon, to create a cross-country microsimulation model, EUROMOD (Atkinson et al. 2002). Other partners, such as Gert Wagner at the Deutsches Institut für Wirtschaftsforschung (DIW),1 provided a link back to the cutting-edge Sfb3 models in Germany in the 1980s (Galler and Wagner 1986). In addition, the early version of the EUROMOD framework was developed together with Herwig Immervoll, who now heads up microsimulation analysis at the OECD (Immervoll and O’Donoghue 2009), although later versions are more flexible
1 German Institute for Economic Research.
and powerful (Sutherland and Figari 2013). Extensions of EUROMOD involved collaborations with Andre Decoster to incorporate modelling consumption and indirect tax (Decoster et al. 2010), and also collaborations with Ugo Colombino to incorporate labour supply (Colombino et al. 2010).
Extending microsimulation modelling to incorporate the spatial dimension, I developed, on returning to Ireland, SMILE, working in partnership with Graham Clarke and Dimitris Ballas of the University of Leeds (Ballas et al. 2005). SMILE has a focus on rural- and agricultural-policy analysis (O’Donoghue et al. 2013). In parallel, my Ph.D. framework, LIAM (O’Donoghue et al. 2009), was generalized to be applied to other countries (Dekkers et al. 2010), which eventually led to the development of a new and faster, more-powerful framework, LIAM2 (de Menten et al. 2014), with the methodologies influencing the development of the UK’s dynamicmicrosimulation model Pensim2, at the Department of Work and Pensions (Edwards 2010; O’Donoghue et al. 2010).
Progress in research is based on building upon the achievements and learnings of others. Much of the learning in microsimulation has been by word of mouth, via interpersonal interactions, or via documentation in conference proceedings or books, many of which are now out of print, such as the excellent Orcutt et al. (1986).
Without personal interactions with leading figures in this fast-growing and relatively novel field, it would have been more challenging for me to develop these models. In essence, I had to rely largely on an oral exchange of knowledge and experience. Across the microsimulation field, much of the knowledge has been transmitted verbally between people on teams, at conferences, and through networks. In developing further, the microsimulation field faces a challenge to find more-effective ways of transferring knowledge. In transferring knowledge, there are two main types of knowledge transfer:
• codified knowledge, where specific knowledge is written down
• tacit knowledge, where more-abstract information may be more difficult to transmit
For much of the period since the foundation of the field, knowledge has been codified mainly through the following forms:
• documentation that aims to facilitate other team members utilizing the models
• published material, mainly books and conference presentations, which may have been non-peer reviewed, had limited coverage, and often went out of print
• documents that may have only been available to those who attended an event and were rarely included in the usual citation indices and searchable databases
• papers published in peer-reviewed formats, which were typically in journals where the focus was on the application rather than the methodology
A significant proportion of the methods used in the field are not formally codified, meaning that new models have had to reinvent the wheel and redevelop existing methods over and over again. Where methods were formally codified, they were often codified in non-peer-reviewed technical notes or discussion papers and thus lack the quality assurance that peer review can help to achieve. Another issue is that publication in a research-centre technical paper or note carries risks associated with the ending of funding, retirement of staff, or the end of the life of a model. Thus, there is a sustainability risk for the field in respect to its core methodological foundations.
A classic example of this is the methodology in relation to alignment, used in dynamic microsimulation models. It is a calibration mechanism used to align simulated totals to external control totals, and has been used since the 1970s. It is, thus, a core methodology within the field. However, there is relatively little documentation or guidance as to how to undertake alignment. Where it exists, it is published in non-peer-reviewed technical papers (Bækgaard 2002) as team-specific internal documentation (Johnson 2001; Morrison 2006), conference papers (Kelly and Percival 2009; Chénard 2000a), or in relatively hard-to-find volumes based on conferences (Neufeld 2000; Chénard 2000b). It should be noted that all these references date from 2000 onward, despite the methodology being used since the 1970s. Most are not peer reviewed and most are hard to find, and, given the dissolution of some of the teams, are impossible to access. One of the first peer-reviewed journal articles that aims to assess the performance of a part of the methodology was only published in 2014 (Li and O’Donoghue 2014). This chapter covers the alignment of only a single variable type. Is it any wonder that the methodology has received serious criticism (Winder 2000)?
It is arguable that the development of a method cannot be trusted until it has been road-tested through publication and rigorous peer review. There is, thus, a need for a literature to be developed to document, test, and provide rigorous quality assurance for the alignment of the many other variables that are found in the literature. The example above cites an issue in relation to one specific aspect of the methodology. This criticism could be extended to many other methods used within the field of microsimulation.
This book is an attempt to codify and describe many of the main techniques utilized in microsimulation modelling, and to present examples of how they are used.
I am grateful both to my many peers in the field of microsimulation and to students that I have learned from and taught over the past twenty-five years. I am also grateful to helpful comments by anonymous referees and to my colleague Mary Ryan for extensive comments on the draft document. I hope this book provides a helpful guide for those wishing to develop models within the field. I would like to acknowledge the understanding and support of Rosaleen and Jude as I prepared this book.
II.
PART III. BEHAVIOURAL MODELS
5.3
5.4
5.5
5.6
7.1
7.2
7.3
7.4
PART IV. DYNAMIC AND SPATIAL MODELS
5.3
5.4
5.5
8.1
8.3
8.4
9.8
10.1
List of Tables
7.1
1 Introduction
1.1 Introduction
Public policy design increasingly expects and relies upon a body of evidence to make decisions. Better evidence can produce better and more focused policies. It can allow for better targeting of resources, improving the cost of achieving a particular policy objective, or improving the effectiveness of a policy for a given resource. Targeted policy interventions require better information on who is affected, how they are affected, and where those who are affected are located. For example, a transfer programme targeted at a group that are generally poor, such as the elderly, may cost more than an instrument that is targeted on the basis of income, and so is targeted specifically at the poor. However, this targeting or means testing may introduce negative incentives. Thus, designing effective policy requires a micro-based unit of analysis, containing information on how a policy will affect individuals differentially.
While there is a large range of methodologies utilized in undertaking evidencebased policy analysis, they can be classified broadly into the following categories:
• Expost analysis (Heckman et al. 1999; Todd 2007; Vedung 2017; Benhassine et al. 2015), focusing on evaluating the impact of a policy after it has been implemented.
• Exante analysis, assessing the potential impact before roll out (Hertin et al. 2009; Figari et al. 2015; De Agostini et al. 2018).
Increasing use is being made of pilot initiatives using randomized experiments and then evaluated using expost methods, for example in the case of the main worldwide pilot projects for childrelated conditional cashtransfer programmes (see Gertler 2004; Fernald et al. 2008; Pearce and Raman 2014; Haskins and Margolis 2014). However, political constraints and/or time or resource constraints frequently do not allow this to take place. Thus exante simulationbased methods are often used for public policy design as they are cheaper, being undertaken on a computer without incurring large piloting Practical Microsimulation Modelling. Cathal O’Donoghue. Oxford University Press. © Cathal O’Donoghue 2021. DOI: 10.1093/oso/9780198852872.003.0001
costs. They are less accurate than expost methods as the structure of behaviour may change in response to policy instrument or they may be based upon historical data. However, the methodology may be the only one possible in many circumstances.
Microsimulation modelling is a potential simulationbased tool with a microunit of analysis that can be used for exante analysis (O’Donoghue 2014). It is a microbased methodology, typically utilizing microdata units of analysis, for example taking surveys or datasets containing microunits such as households, individuals, firms, and farms, etc. It is a simulationbased methodology that utilizes computer programs to simulate public policy and economic or social changes on the micropopulation of interest. While microsimulation models have taken firms (Eliasson 1991; Buslei et al. 2014) or farms (O’Donoghue 2017) as the microunit of analysis, most have carried out analysis at the level of individuals or households (see Mot 1992; Sutherland and Figari 2013)
As a research field, microsimulation has its roots in the work of Guy Orcutt (1957, 1961). However, it was only the advent of the personal computer in the 1980s and the availability of microdata that have allowed the field to develop. Whether formally defined as microsimulation modelling or not, microbased, exante simulationbased analysis is now used extensively around the world for policy analysis and design.
There have been a number of survey articles written such as Merz (1991, 1994), Mot (1992), Martini and Trivellato (1997), Bourguignon and Spadaro (2006), Dekkers and van Leeuwen (2010), and Anderson and Hicks (2011). Generally, Sutherland (1995) covered static models; Klevmarken (1997) behavioural models; O’Donoghue (2001), Zaidi and Rake (2001), Spielauer (2007), and Li and O’Donoghue (2013) dynamic models; Rahman and Harding (2016), Rahman et al. (2010), Hermes and Poulsen (2012), Tanton and Edwards (2013), Tanton (2014), and O’Donoghue et al. (2014) spatial models; Creedy and Duncan (2002), Creedy and Kalb (2005), and Bargain and Peichl (2013) labour supply models; Figari and Tasseva (2013) a special issue on the crosscountry EUROMOD model; Brown (2011) health models; and Ahmed and O’Donoghue (2007), Cockburn et al. (2010), and Bourguignon et al. (2010) covered macromicro models. The O’Donoghue (2014) handbook brings together developments across a variety of different areas. Given the growth in microsimulation over the past twenty years, there is a need for a text book to assimilate this literature and describe the development and implementation of current practice in the microsimulation field. Public policy is broad, with many objectives and associated targets. Microsimulation modelling can in principle be applied to assess the micro impact of many policy areas, subject to data availability and to the capacity
to quantify the impact of the policy. This book will focus primarily on policies associated with the distribution of income such as poverty, income inequality, and labour supply incentives. These are the areas in which the methodology has seen most use over time. However, there are also many other areas in which the methodology has been widely used, particularly in the areas of transport (Miller 2014), health (Schofield et al. 2014), urban planning (Waddell et al. 2003), and farmlevel modelling (O’Donoghue 2017; Shrestha et al. 2016).
Microsimulation models can be produced in different programming environments. Relatively simple models can be programmed in Microsoft EXCEL, while more sophisticated models use statistical software such as SAS or Stata, or programming languages such as C++, VB, and Java (Hancock 1997). There are no specific software packages for undertaking microsimulation, but a number of frameworks have been used to build models for different purposes, for example the EUROMOD (Immervoll and O’Donoghue 2009), MODGEN (Spielauer 2011), and LIAM2 (De Menten et al. 2014) frameworks.
Other modelling methods, such as computable general equilibrium models (CGE) (De Melo 1988; Van Ruijven et al. 2015), overlapping generations models (OGM) (Lambrecht et al. 2005; Bommier and Lee 2003), or agentbased models (Tesfatsion and Judd 2006; Gatti et al. 2018), incorporate behaviour in a more detailed or consistent way than microsimulation models, but typically do not have the same heterogeneity of population or detail in relation to policy. Linking these models with microsimulation models can generate some of the advantages of both methods, illustrated by attempts to link more detailed behavioural models such as CGE (Cockburn et al. 2014) with microsimulation models.
1.1.1 Complexity and Microsimulation Models
As a modelling framework, microsimulation modelling is a mechanism of abstracting from reality to help us understand complexity better. Figure 1.1 outlines potential sources of complexity in a static, singletimeperiod microsimulation model.
In the context of policy design and evaluation, complexity can take the form of:
• population structure
• behavioural response to the policy
• policy structure
