THE OXFORD HANDBOOK OF HEALTH ECONOMICS THE OXFORD HANDBOOK OF HEALTH ECONOMICS Edited by SHERRY GLIED and PETER C SMITH 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 © Oxford University Press 2011 The moral rights of the authors have been asserted First published in 2011 First published in paperback 2013 Impression: 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 British Library Cataloguing in Publication Data Data available Library of Congress Cataloging in Publication Data Data available ISBN 978–0–19–923882–8 (hbk) ISBN 978–0–19–967540–1 (pbk) Printed in Great Britain on acid-free paper by Ashford Colour Press Ltd, Gosport, Hampshire 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 ACKNOWLEDGEMENTS THE editors would like to thank all those who contributed their expertise to this volume At Oxford University Press, the handbook was commissioned by Sarah Caro, who guided the early development of the volume, and brought to a successful conclusion by Georgia Pinteau We also greatly appreciated the help of the publisher’s editorial team, which included Chris Champion, Emma Lambert, Rachel Platt, and Aimee Wright Great thanks are due to Vanessa Windass, who provided unfailing secretarial support throughout the project, and to Dahlia Rivera Finally, we should like to thank the authors, whose scholarship is evident throughout these pages, and who responded to editorial suggestions with great wisdom and patience CONTENTS List of Figures List of Tables List of Boxes List of Contributors Introduction SHERRY GLIED AND PETER C SMITH Health Systems in Industrialized Countries BIANCA K FROGNER, PETER S HUSSEY, AND GERARD F ANDERSON Health Systems in Low- and Middle-income Countries ANNE MILLS The Political Economy of Health Care CAROLYN HUGHES TUOHY AND SHERRY GLIED The Promise of Health: Evidence of the Impact of Health on Income and Well-Being WILLIAM JACK Health Production KRISTIAN BOLIN Socioeconomic Status and Health: Dimensions and Mechanisms DAVID M CUTLER, ADRIANA LLERAS-MUNEY, AND TOM VOGL Determinants of Health in Childhood MICHAEL BAKER AND MARK STABILE Economics of Infectious Diseases RAMANAN LAXMINARAYAN AND ANUP MALANI 10 Economics of Health Behaviors and Addictions: Contemporary Issues and Policy Implications DONALD S KENKEL AND JODY SINDELAR 11 Economics and Mental Health: An International Perspective RICHARD G FRANK 12 Public-Sector Health Care Financing ÅKE BLOMQVIST 13 Voluntary Private Health Insurance PETER ZWEIFEL 14 Health Care Cost Growth MICHAEL E CHERNEW AND DUSTIN MAY 15 User Charges ERIK SCHOKKAERT AND CARINE VAN DE VOORDE 16 Insurance and the Demand for Medical Care MARK V PAULY 17 Guaranteed Access to Affordable Coverage in Individual Health Insurance Markets WYNAND P M M VAN DE VEN AND FREDERIK T SCHUT 18 Managed Care LAURENCE BAKER 19 Hospitals: Teaming Up PEDRO PITA BARROS AND PAU OLIVELLA 20 Primary Care ANTHONY SCOTT AND STEPHEN JAN 21 The Global Health Workforce TILL BÄRNIGHAUSEN AND DAVID E BLOOM 22 The Economics of the Biopharmaceutical Industry PATRICIA M DANZON 23 Disease Prevention, Health Care, and Economics JANE HALL 24 Long-term Care JOSE-LUIS FERNANDEZ, JULIEN FORDER, AND MARTIN KNAPP 25 Physician Agency and Payment for Primary Medical Care THOMAS G MCGUIRE 26 Provider Payment and Incentives JON B CHRISTIANSON AND DOUGLAS CONRAD 27 Non-price Rationing and Waiting Times TOR IVERSEN AND LUIGI SICILIANI 28 Increasing Competition Between Providers in Health Care Markets: The Economic Evidence CAROL PROPPER AND GEORGE LECKIE 29 Measuring Organizational Performance JAMES F BURGESS JR AND ANDREW STREET 30 Health System Productivity JACK E TRIPLETT 31 The Methods of Cost-effectiveness Analysis to Inform Decisions about the Use of Health Care Interventions and Programs SIMON WALKER, MARK SCULPHER, AND MIKE DRUMMOND 32 Analyzing Uncertainty in Cost-effectiveness for Decision-making SUSAN GRIFFIN AND KARL CLAXTON 33 Health Utility Measurement DONNA ROWEN AND JOHN BRAZIER 34 Concepts of Equity and Fairness in Health and Health Care JAN ABEL OLSEN 35 Measuring Inequality and Inequity in Health and Health Care EDDY VAN DOORSLAER AND TOM VAN OURTI 36 Inter-generational Aspects of Health Care LOUISE SHEINER 37 Econometric Evaluation of Health Policies ANDREW M JONES AND NIGEL RICE 38 Health Economics and Policy: The Challenges of Proselytizing ALAN MAYNARD AND KAREN BLOOR Index LIST OF F IGURES 2.1 Health care sector share of total health care spending in the median OECD country, 1970–2005 2.2 Decomposition of average annual growth in health spending into growth in GDP, aging, and excess in OECD countries, 1970–2005 3.1 The health system 3.2 Out-of-pocket share of total health expenditure in relation to GDP per capita 3.3 Median coverage levels for priority maternal, neonatal, and child health interventions (68 priority countries) 3.4 Use of public and private services 3.5 Volatility of external funding 3.6 External funding for maternal and neonatal health in relation to need 4.1 Government health expenditures as a share of all health expenditures and as a share of all government outlays, 2005 5.1 Life expectancy per capital GDP (US dollars) 6.1 Illustration of the productivity of health capital 6.2 Illustration of the demand for health capital 6.3 Illustration of the effect of age on the demanded amount of health capital 6.4 Illustration of the effect of a wage rate increase 6.5 Illustration of the effect of an increase in educational capital on the demanded amount of health capital 7.1 Education and mortality among adults over 40, US and Europe 7.2a Education and mortality, US adults over 25 7.2b Education and self-reported health, US adults over 25 7.3a Income and mortality, US adults over 25 7.3b Income and self-reported health, US adults over 25 7.4 Occupation and mortality, US adults ages 25–65 7.5a Race and mortality, US adults over 25 7.5b Race and self-reported health, US adults over 25 11.1 Managed care rationing by shadow prices 11.2 Budgets and incentives for public mental health services 13.1 Optimal wealth levels depending on the state of health 14.1 Cost growth in OECD countries 14.2 Total health expenditures as a share of gross domestic product, 1960–2004 15.1 Optimality of user charges as a revenue-raising device 17.1 Three modalities of organizing the payment flows of a subsidy system 18.1 Characteristics of stereotypical types of health insurance plans 21.1 Quotients of urban nurse-to-population ratios divided by rural nurse-to-population ratios 21.2 Quotients of urban physician-to-population ratios divided by rural physician-to-population ratios 21.3 Estimates of health worker emigration rates in sub-Saharan African countries, 2004 24.1 Projected changes in employment (% change of employed people aged 15–64 between 2003 and 2050), EU 25 24.2 Changes in the targeting of community care services for older people by sector of provision, 1993–2008 25.1 Physician preferences and choice of treatment in mixed payment systems 26.1a Panel A—marginal revenue (MR), marginal cost (MC) of quantity 26.1b Panel B—marginal revenue (MR), marginal cost (MC) of quality 29.1 Production frontier: data envelopment analysis 29.2 Production frontier: stochastic frontier analysis 31.1 Assessing the cost-effectiveness of an intervention given an objective of maximizing health subject to a fixed budget 31.2 The incremental cost-effectiveness plane 31.3 Top right quadrant of the cost-effectiveness plane for options Z, W, and Y as defined in Tables 31.1 and 31.2 32.1 Incremental cost-effectiveness plane 32.2 Cost-effectiveness acceptability curves for three mutually exclusive interventions 32.3 Cost-effectiveness acceptability frontier and population EVPI for three mutually exclusive alternatives 33.1 Standard gamble for a chronic health state valued as better than dead 33.2 Time trade-off for a chronic health state valued as better than dead 33.3 Visual analog scale 33.4 Observed and predicted EQ-5D scores using a variety of models mapping the SF-36 and SF12 onto EQ-5D 33.5 Time trade-off for a chronic health state valued as worse than dead 33.6 Predicted EQ-5D health state utility values using the standard and episodic RUM model 33.7 “Lead time” time trade-off for a health state valued as better than dead 33.8 “Lead time” time trade-off for a health state valued as worse than dead 34.1 The determinants of ill health 34.2 Equality, efficiency, and trade-offs 34.3 A more general health frontier 35.1 Two hypothetical concentration curves 35.2 Horizontal inequity indices for probability of a specialist visit, by country (with 95% confidence interval) 35.3 Decomposition of inequity in the probability of a specialist visit (excluding need contributions) 35.4 Short-run (SR) versus long-run (LR) “conservative” inequity for number of GP visits, by country 35.5 Short-run (SR) versus long-run (LR) “conservative” inequity for number of specialist visits, by country 36.1 Health spending by age group, US, 2004 36.2 Distribution of spending by age in Canada contributions earn less than they would if they were invested The loss from the system experienced by later cohorts is the product of their contributions and the difference between r and (1+n)(1+g)–1 The larger are the benefits paid to earlier cohorts, the larger are the payroll contributions and the larger the net present value loss for later cohorts 26 Medicare is separated into three parts—Part A, which mostly covers hospital care, Part B, which finances physician expenses, and the new part D, which covers prescription drugs Part A is funded by payroll taxes, whereas Parts B and D are funded by general revenues Cutler and Sheiner used the age distribution of income taxes to allocate the burden of the general revenues required to fund Medicare Part B 27 This was the Medicare Trustees’ basic assumption at the time the research took place 28 If per capita health spending is growing at a rate z percentage points faster than per capita GDP, then, under a pure pay-as-you-go system, the rate of return is (1+n)(1+g)(1+z), using the terminology from above To sustain this rate of return requires tax payments also be increasing at a rate faster than per capita GDP Eventually, these taxes become too burdensome (in the limit, they comprise all of wages) and then rates of return must fall back to the sustainable level of (1+n)(1+g) 29 Cutler and Sheiner assume that benefits financed by general revenues are fully paid for each year, but that the payroll taxes used to finance Medicare Part A are constant, as under current law 30 As noted, in keeping with the assumption used by the Medicare Trustees at the time, Cutler and Sheiner assumed that per beneficiary Medicare spending would slow to the rate of per capita GDP after twenty years Since then, the assumptions used by the Trustees have changed significantly In particular, the Trustees now assume that the growth rate of per capita Medicare spending slows to the growth rate of per capita GDP only very slowly, so that by the end of the eighty-year projection period, per capita Medicare spending is still increasing a little faster than per capita GDP In addition, a prescription drug benefit was introduced in 2006; according to the Trustees, the Medicare drug benefit raises Medicare outlays by about 15 percent in 2008 and by about 20 percent by 2030 (CMS Report 2008) The rates of return from Medicare under these new assumptions have not been analyzed However, the implications are clear The combination of the different assumptions about the trajectory of health spending and the enactment of the prescription drug benefit will result in significantly higher rates of return for all cohorts save the ones turning 65 near the end of the projection period—that is for all cohorts born before 2015 Of course, the faster pace of assumed growth would also require a more substantial tax increase or benefit cut to make the system sustainable * We gratefully acknowledge funding from the Economic and Social Research Council (ESRC) under the Large Grant Scheme, reference RES-060-25-0045 We are also grateful to Anirban Basu, Bill Greene, Rodrigo Moreno Serra, John Mullahy, Karen Mumford, Silvana Robone, Pedro Rosa Dias, Jo Swaffield, Ranjeeta Thomas, and Pravin Trivedi for their comments on an earlier version In the context of development economics, Banerjee and Duflo (2008) review the advantages of randomized experiments, which allow the assignment of treatment to be isolated and controlled by researchers, in providing internally valid estimates of effectiveness They also seek to address some of the common criticisms of randomized experiments: whether the results of an experiment are dependent on the specific environment where it is carried out, limiting their replicability and generalizability to other contexts; whether there are problems with compliance in experimental studies; whether randomization itself may affect outcomes (for example, through Hawthorne effects); whether experiments only reflect partial equilibrium effects and fail to capture general equilibrium or spillover effects that may occur when policies are implemented on a larger scale; and issues that arise when there is heterogeneity in treatment effects The drawbacks of a mechanical reliance on the experimental approach, and of “quasi-experimental” approaches that use instrumental variables, with instruments selected to mimic a randomized experiment rather than being drawn from a structural model, are reviewed by Deaton (2008) He is critical of the use of these methods to evaluate projects per se and favours their use as tools to aid our understanding of the underlying theoretical mechanisms that drive behaviour This explains the prominence of statistical decision analysis in the recent economic literature on health technology assessment (see e.g Claxton 1999) which has parallels with the work of Manski (2005) in the general program evaluation literature However, it is notable that applications of one the most commonly adopted approaches in health economics—difference-in-differences —rely on a heavily parametric approach, using linear models (see Jones 2009) In this sense the evaluation problem can be seen as a particular brand of missing data problem and many of the methods used to evaluate treatment effects, such as those based on propensity scores, are used in that more general context as well Common support requires that for each level of p(x), the probability of observing a non-treated individual is positive Accordingly, the PATT is identified by restricting attention to comparative non-treated individuals that fall within the support of the propensity score distribution of the treatment group In the case if the PATE the required overlap condition is stronger: < P(d = | x) < Hahn (1998) established that the value of conditioning on the propensity score, rather than on the elements of x directly, stems from the reduction in dimensionality rather than a gain in efficiency for semiparametric estimators In practice matching on the propensity score may be combined with exact matching on specific covariates or matching within subgroups of the sample defined by the covariates The Mahalanobis metric, which scales differences in x by the inverse of their covariance matrix, is often used for exact matching of covariates Note that this argument applies for the estimation of average treatment effects, that rely on additivity, but does not apply to other measures such as quantile treatment effects The strong assumptions that are often required to achieve identification of point estimates for treatment effects, in the presence of selection on unobservables, lead Manski to focus on partial identification and to propose estimates of the bounds for treatment effects (e.g Manski 1990) 10 If any of the set of exogenous explanatory variables, x, are omitted from the first stage, then this might induce correlation between the omitted variables and the second stage residuals, potentially leading to an inconsistent estimator of the treatment effect 11 The definition of the LATE relates to instruments that are monotonically related to treatment assignment 12 Excluding one individual to represent the baseline case against which the effects of others can be contrasted 13 Inferences concerning the estimate of τ depend on the assumptions made about the error term ε and it may not be reasonable to it assume serial independence Bertrand, Duflo, and Mullainthan (2004) have suggested that standard errors should be adjusted to allow for clustering within individuals in applications of DD ... Economics Anne Mills is Professor of Health Economics and Policy, Head of the Faculty of Public Health and Policy, and Director of the Health Economics and Financing Programme at the London School of. .. publishes in the area of micro-econometrics and health economics, with emphasis on the determinants of health, the economics of addiction, and socioeconomic inequalities in health and health care Donald... University of Chicago His research focuses on the economics of health promotion and disease prevention Martin Knapp is Professor of Social Policy at the London School of Economics and Professor of Health