A practitioners guide to state and local population projections, stanley k smith, jeff tayman, david a swanson, 2013 998

The Springer Series on Demographic Methods and Population Analysis 37 Stanley K Smith Jeff Tayman David A Swanson A Practitioner's Guide to State and Local Population Projections A Practitioner’s Guide to State and Local Population Projections THE SPRINGER SERIES ON DEMOGRAPHIC METHODS AND POPULATION ANALYSIS Series Editor KENNETH C LAND Duke University In recent decades, there has been a rapid development of demographic models and methods and an explosive growth in the range of applications of population analysis This series seeks to provide a publication outlet both for high-quality textual and expository books on modern techniques of demographic analysis and for works that present exemplary applications of such techniques to various aspects of population analysis Topics appropriate for the series include: • • • • • • • • • • • General demographic methods Techniques of standardization Life table models and methods Multistate and multiregional life tables, analyses and projections Demographic aspects of biostatistics and epidemiology Stable population theory and its extensions Methods of indirect estimation Stochastic population models Event history analysis, duration analysis, and hazard regression models Demographic projection methods and population forecasts Techniques of applied demographic analysis, regional and local population estimates and projections • Methods of estimation and projection for business and health care applications • Methods and estimates for unique populations such as schools and students Volumes in the series are of interest to researchers, professionals, and students in demography, sociology, economics, statistics, geography and regional science, public health and health care management, epidemiology, biostatistics, actuarial science, business, and related fields For further volumes: http://www.springer.com/series/6449 Stanley K Smith • Jeff Tayman David A Swanson A Practitioner’s Guide to State and Local Population Projections Stanley K Smith Bureau of Economic and Business Research University of Florida Gainesville, FL, USA Jeff Tayman Economics Department University of California-San Diego San Diego, CA, USA David A Swanson Department of Sociology University of California Riverside Riverside, CA, USA ISSN 1389-6784 ISBN 978-94-007-7550-3 ISBN 978-94-007-7551-0 (eBook) DOI 10.1007/978-94-007-7551-0 Springer Dordrecht Heidelberg New York London Library of Congress Control Number: 2013956645 © Springer Science+Business Media Dordrecht 2013 This work is subject to copyright All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed Exempted from this legal reservation are brief excerpts in connection with reviews or scholarly analysis or material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work Duplication of this publication or parts thereof is permitted only under the provisions of the Copyright Law of the Publisher’s location, in its current version, and permission for use must always be obtained from Springer Permissions for use may be obtained through RightsLink at the Copyright Clearance Center Violations are liable to prosecution under the respective Copyright Law The use of general descriptive names, registered names, trademarks, service marks, etc in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use While the advice and information in this book are believed to be true and accurate at the date of publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be made The publisher makes no warranty, express or implied, with respect to the material contained herein Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com) Foreword A large part of our daily lives is governed by numbers How many hours of sleep did I get last night? How many unread messages are queued up in my inbox? How many “friends” I have on Facebook? What’s the upcoming Powerball payoff? How’s my cholesterol count doing? Can I recall my daughter’s phone number, my granddaughter’s birthday? Numbers such as these encompass portions of our personal and shared social reality They roll around in our head, and they are part of what determine our mood, our behavior, well-being, worries, and activity constraints Population projections also present us with numbers But these are numbers of a very different nature Rather than simply reflecting a social reality (and associated beliefs and behaviors), they serve to create a reality based on anticipation—a reality unwitnessed, unobserved, and largely unknown Yet, on the basis of such numbers, schools are built (or closed), roads are widened, airport terminals expanded, municipal services extended, and marketing strategies altered So this book is about the second kind of number, the sort leading to anticipatory behaviors and, occasionally, preemptive actions It is the applied demographer’s difficult role to creatively deploy the data, tools, and perspectives of the population sciences to carry out these tasks not only ethically and transparently but with an experienced and disciplined hand This book provides a marvelously clear, well-organized, and comprehensive blueprint for understanding and competently performing this role The authors are seasoned applied demographic practitioners They have individually and collaboratively contributed mightily to the demographic literature The book is intellectually solid, methodologically encompassing, and—while retaining the historically interesting material—firmly contemporary and up to date Early in my own career, one of my mentors rhetorically asked, “Why we make population projections? They always turn out to be wrong, so why we persist?” After a brief excursion through the standard reasons for why projections are useful, he added, “Probably the most important reason for engaging in this enterprise is so that we later know what to be surprised about.” Wonderfully said! The point is that we live in a world of frenetic change We are so habituated to mindlessly accommodating to this change that we rarely pause long enough to say “Wow!” and to reflect on what we thought, just a few years back, our present reality v vi Foreword might look like Projections and forecasts prepared yesterday permit us to that today Today’s projections will serve that potentially chastening purpose tomorrow The overall narrative and thoroughly developed methodologies in A Practitioner’s Guide to State and Local Population Projections will, of course, not eliminate all of tomorrow’s surprises One can indeed hope, however, that practitioners who wisely select to use this book to guide their own demographic pursuits will benefit from the authors’ skillful verve, their richly detailed methodological coverage, and the numerous concrete examples presented throughout the material to minimize the number and magnitude of future surprises This valuable compendium presents methods that are tried and tested alongside those that are recent and innovative The book revisits and updates most of the topics treated in the authors’ earlier book, State and Local Population Projections: Methodology and Analysis However, the current book should not be understood as a revised edition Fresh attention is given to emerging methodological approaches in small-area population forecasting (projecting) and, in particular, to new data resources that have fundamentally altered the information content of forecasting models The material is treated with originality and conviction and benefits immensely from real-life illustrations drawn from the authors’ own work With A Practitioner’s Guide to State and Local Population Projections, Smith, Tayman, and Swanson have again secured their leading place as careful, practical, and solidly competent applied demographic scholars University of North Carolina at Chapel Hill June 2013 Paul R Voss Preface A lot has happened since we published State and Local Population Projections: Methodology and Analysis in 2001 Smart phones, electronic tablets, and the Cloud have given us access to virtually endless sources of information, no matter where we are Improvements in technology, software, and computing hardware have expanded the way we access, store, and analyze information Facebook, Twitter, and other social media sites have changed the way we communicate Globalization has altered the world’s economic system and 9/11 changed almost everything, from international relations to the way we board airplanes A lot has happened in the field of applied demography as well Data sources have proliferated, methods have advanced, computing capabilities have mushroomed, and new research has been published One major change is that the decennial census no longer includes a long form collecting detailed socioeconomic, demographic, and housing information from a sample of census respondents This information is now collected in the American Community Survey (ACS), which differs in several ways from the census long form Particularly important for the production of population projections is that the ACS collects data continuously rather than once per decade, is based on a smaller sample size, uses different residence rules, and measures migration over a 1-year rather than a 5-year period These and other changes have convinced us that a new book on state and local population projections is needed This new book retains and updates much of the material included in our previous book, but covers a number of new topics as well We present a detailed discussion of the differences between ACS migration data and the migration data collected in the decennial census, paying particular attention to how these differences affect the construction of cohort-component projections We provide an illustration of how to use ACS migration data to project a county population We add a new chapter on projections of population-related variables such as households, school enrollment, labor force participation, and persons with disabilities We expand our discussion of microsimulation models, scenario analysis, special populations, international migration, and the benefits of combining vii viii Preface projections from several different models Throughout the book, we incorporate research findings that have appeared in the literature since the publication of our previous book As before, we pay particular attention to problems encountered when making projections for small areas (e.g., counties and subcounty areas) We describe a number of data sources and projection methods, focusing on those that are most accessible and can be used in a variety of circumstances We discuss the strengths and weaknesses of each and provide our thoughts on which are most useful for particular purposes We include many examples and illustrations, as well as equations and verbal descriptions, in an attempt to present the material as clearly as possible A number of methods, data sources, and application techniques can be used for constructing state and local population projections Deciding which ones to include— and how to present them—was not an easy task We wanted the book to be comprehensive but not long-winded, technically precise but not overly mathematical, clearly written but not simplistic We wanted it to be useful to analysts with a strong background in demography yet accessible to those with little or no demographic training Most important, we wanted it to provide practical guidance to demographers, planners, market analysts, and others called upon to construct or evaluate state and local population projections in real-world settings The reader will have to decide whether we have succeeded in accomplishing these often-conflicting goals We thank Paul Voss for writing the foreword to this book It would be impossible to find a person more qualified than Paul, given his numerous and important contributions to the field of applied demography We also thank Evelien Bakker and Bernadette Deelen-Mans for shepherding the book through Springer’s production process; their assistance was invaluable Above all, we express our gratitude to our wives—Rita, Melinda, and Rita—for their love, encouragement, and patience as we worked on this book Contents Rationale, Terminology, Scope 1.1 What Is a Population Projection? 1.1.1 Projections, Forecasts, Estimates 1.1.2 Alternative Approaches 1.2 Why Make Population Projections? 1.2.1 Roles of Projections 1.2.2 Projections and Decision Making 1.2.3 Forecasting and Planning 1.3 How Can This Book Help? 1.3.1 Objectives 1.3.2 Geographic Focus 1.3.3 Coverage 1.3.4 Target Audience References 2 6 10 11 11 14 15 17 17 Fundamentals of Population Analysis 2.1 Demographic Concepts 2.1.1 Size 2.1.2 Distribution 2.1.3 Composition 2.1.4 Change 2.2 Components of Change 2.2.1 Mortality 2.2.2 Fertility 2.2.3 Migration 2.2.4 Demographic Balancing Equation 2.3 Statistical Measures 19 19 19 22 23 25 27 28 28 29 29 32 ix 384 14 A Practical Guide to Small-Area Projections The course of future population change is also uncertain The degree of uncertainty grows as the projection horizon becomes longer and as the size of the population becomes smaller We can be much more confident of a 1-year forecast than a 20-year forecast Similarly, we can be more certain of the future size of the U.S population than the future size of the population living in Portland, Maine The errors reported in Chap 13 illustrate the uncertainty inherent in population forecasts We believe it is important to provide data users with some indication of uncertainty This can be done in several ways One is to construct a range of projections based on two or more methods or different specifications of a particular method For example, projections might be made using several trend extrapolation methods and/or different base periods for each one A more common approach is to produce several sets of cohort-component projections based on different combinations of assumptions For example, fertility rates could be projected to rise by 10%, fall by 10%, or remain constant Migration rates could be based on data from the last years, years, or 10 years The primary benefit of producing a range of projections is that it shows the populations stemming from different models, techniques, or sets of assumptions The primary limitation is that it does not provide an explicit measure of uncertainty How likely is it that the future population will fall within the range suggested by two alternative projections? How likely is it that any particular projection will provide an accurate forecast of future population change? These questions cannot be answered simply by producing a range An explicit measure of uncertainty can be given by constructing prediction intervals to accompany population forecasts (see Chap 13) These intervals can be based on specific models of population growth or on empirical analyses of past forecast errors Model-based intervals are difficult to produce and are subject to a variety of specification errors Empirically-based intervals require the collection of a large amount of historical data Both are valid only to the extent to which future error distributions are similar to past or simulated distributions In spite of these problems, prediction intervals offer one major advantage over a range of projections: They provide an explicit measure of the uncertainty surrounding future population growth Forecast uncertainty can also be assessed by comparing more broadly defined projection scenarios For example, scenarios can be based on alternative assumptions regarding land use patterns; the transportation system; and housing, economic, fiscal, and environmental policies Under this approach, uncertainty is evaluated by considering various policy options; it is not intended to show a specific range of values or a precise numerical estimate for a particular area Another way to provide data users with some indication of uncertainty is to construct tables summarizing errors from previous forecasts for the area to be projected or for areas with similar characteristics Although this approach does not provide an explicit range or prediction interval, it does provide an assessment of past performance and—by extension—a basis for predicting future performance 14.3 Review and Document the Results 385 Of course, it is important to remember that—as buried in the small print of mutual fund advertisements—past performance is no guarantee of future performance In some instances data users may be better off using a high or low projection rather than the forecast or “most likely” projection For example, suppose that the cost to a city of building too large a sewer system is relatively small, but the cost of building too small a system is very large To reduce risk it may be advisable to use a projection from the high series or at the high end of the prediction interval for planning the size of the system Estimates of the cost of being wrong may play an important role in the choice of the projection to be used for any particular purpose Measures of uncertainty help the data user make these choices Population projections used as forecasts are subject to error, especially for small areas, areas that have been growing or declining rapidly, and for long forecast horizons These errors are caused by our inability to correctly predict the future course of mortality, fertility, and migration We believe it is important to convey this information to the data user Although it may be disappointing, information on potential errors will give data users a more realistic view of the future and help them plan more effectively for the uncertainty inherent in population projections 14.3 Review and Document the Results The analyst may believe the job is finished once the steps described above have been completed That would be a mistake At this point, the projections should be viewed as strictly preliminary Before they are finalized they must be thoroughly reviewed and evaluated Are the results plausible? Do they make sense given historical population trends in the area and projected trends in other areas? Are they consistent with the area’s demographic characteristics and economic conditions? It is possible that there were flaws in the original projection methodology and assumptions, that errors were made in data entry or programming, or that something important was overlooked A review and evaluation of the results will often uncover such problems As the final step in the projection process, the entire methodology must be documented, describing all data sources, models, techniques, and assumptions 14.3.1 Internal Review By internal review, we mean an examination of the results by the person or agency producing the projections This can be done in a variety of ways (Dion 2012; Murdock et al 1991) We suggest several that we have found to be particularly helpful Suppose we are reviewing a set of county projections It is useful to observe historical population trends and to compare past changes with projected changes 386 14 A Practical Guide to Small-Area Projections Table 14.1 Average annual population change for Florida and selected counties, 1980–2010 Place 1980–1985 1985–1990 1990–1995 1995–2000 2000–2005 2005–2010 Broward 21,307 26,139 34,644 38,862 23,828 1,182 Duval 9,615 10,779 9,543 11,639 9,853 7,224 Marion 7,291 7,178 6,040 6,777 8,274 6,202 Pinellas 14,552 10,073 7,222 6,745 2,295 À3,286 Sumter 670 791 888 3,466 3,163 4,852 Union 87 À70 457 181 273 145 Florida 315,075 323,118 279,612 329,368 359,066 204,631 Source: Bureau of Economic and Business Research, University of Florida, unpublished data Table 14.1 shows average annual population changes for Florida and several of its counties from 1980 to 2010 Duval and Marion exhibited fairly stable changes over the entire time period, albeit with a bit of a decline in 2005–2010 due to the severe economic recession near the end of the decade Broward exhibited changes that increased steadily during the 1980s and 1990s but declined rapidly thereafter Sumter exhibited changes that rose rapidly over time and Pinellas exhibited changes that declined rapidly, becoming negative for 2005–2010 Union exhibited substantial volatility over the entire time period Rapidly changing trends is a red flag warning the analyst to investigate the accuracy of the input data If no errors are found, potential causes of the changes must be considered There are logical explanations for several of the trends shown in Table 14.1 The volatility in Union County’s population was caused by fluctuations in its large prison population Increases in Sumter County’s prison population contributed its growth in the late 1990s, but the main cause of the large population increase since 2000 was the development of a huge retirement community Pinellas County is geographically small and densely populated; the steady decline in its population growth since 1980 was caused primarily by the growing scarcity of open space for further expansion The severe recession in 2007–2009 slowed population growth almost everywhere in the state Figuring out the causes of recent trends will help the analyst refine the methodology and perhaps revise the assumptions used in creating the projections Analyzing historical trends gives the analyst a basis for evaluating the plausibility of the projections Are projected changes consistent with historical changes? If not, why not? If a logical explanation cannot be found, it may be a tip-off that an error was made in choosing the projection techniques, developing assumptions, or writing computer programs Inspecting patterns of population change over the projection horizon also provides helpful clues Do the projected changes become larger, smaller, or remain about the same as the horizon becomes longer? Is there a logical explanation for this pattern? If not, this is another red flag that must be investigated Similar tables could be constructed showing percent changes rather than numeric changes, or showing county population as a share of state population These alternative forms provide different perspectives for viewing population change and judging plausibility The specific form doesn’t matter What matters 14.3 Review and Document the Results 387 Table 14.2 Percent distribution of the population by age for Florida, and selected counties, 1980–2010 Place Florida Age