Defining and Using Residential Submarkets in Planning Work Clifford A. Lipscomb, Ph.D. Director of Economic Research Greenfield Advisors, LLC Seattle, WA and Atlanta, GA United States of America Greenfield Advisors • 37-year-old firm headquartered in Seattle, WA, USA • Real estate valuation, economic research, survey design and administration, specialty in complex valuation issues (i.e. we don’t do appraisals for banks/lenders) • Most of our work is litigation support • Current cool project: patent infringement The Big Picture • People and households are heterogeneous (different) • How different are they? – Socioeconomic – Racial – Demographic – Others • How can we measure the differences? Are they “systematic”? Why is This Important? • Price prediction (submarkets improve model accuracy) • Formulation of market strategy • Understanding housing market structure • Improving lenders’ and investors’ ability to price risk associated with homeownership • Public policy implications – Policy tailored to one “type” may not be best – Policy inertia (is a mid-course correction possible after a policy is implemented?) Terminology • What is a market? • What is a submarket? • How can we define submarkets? – Housing stock (similar packages of housing services) – Geography (traditional neighborhoods) – Household characteristics (data intensive) – Surveys that ask about who are your “peers” – good for comparative studies – Hybrid methods Rest of the Presentation • Focus on an Atlanta neighborhood • Talk about external and internal factors affecting the neighborhood • Discuss how a change in land use can affect the neighborhood Home Park Sales Price Distribution North Home Park South Home Park Home Park green space Less than $150K $150K – 199,999 $200K – 249,999 $250K – 299,999 $300K + Determining Submarkets • Motivation – assumptions in the literature • Publically available data was limited (Census tract block group was smallest unit available) • Houses are the unit of analysis, so need that level of detail in demographics • Differences between renters and owners • Door-to-door survey effort • 51% response rate Empirical Model • Round 1: Cluster analysis establishes groups • Round 2: Refines groups into “types” based on a variation of linear regression model (SUR) • Houses are sorted into types based on the appraised value that minimizes error • Determines the number of types without researcher pre-determination! Because what if the researcher draws an arbitrary boundary… Original Household Types Type A Type B Type C Submarkets • A: Undergraduate student renters • B: Other student renters, young professionals, and young married couples • C: Owners and graduate students • Note: Recent research has tightened the distinctions between submarkets using different econometric estimators (Belasco, Farmer, and Lipscomb 2012) Dynamic Issues • What happens to neighborhood if you put in a pocket park? • Simulation results – Simulation 1: if preference for park access stays same – Simulation 2: if preference for park is ½ of current estimate Location of New Pocket Park Type A Type B Type C What Happens After Re-sorting? • Models predict that mix of residents will change by 30% as a result of pocket park • Student renters are “forced out” as new owner-occupiers enter the neighborhood • Change in amenity mix = change in occupant • Did pocket park simply accelerate resident mix that was going to happen anyway? Types After Pocket Park Type A Type B Type C Type A Type B Type C Implications • Policymakers and planners need to plan with preference heterogeneity in mind • A more granular level of data needed to complete comprehensive plans • E.g. new MARTA rail stop will be used by what “type” of households? • E.g. what “types” are attracted to TODs? Implications (2) • Balancing housing affordability with housing construction (micro-apartments targeting urban professionals living alone) • Planners can influence the “types” of residents attracted to a neighborhood – influence can be latent or manifest • Planners plan for change and can influence that change Summary • Economists and planners often use data at one geographic scale when analyzing phenomena at a different scale • Make sure statistical methods are an empirical translation of your theory • Beware of post hoc ergo propter hoc fallacy (“after this, therefore because of this”) • All methods have limitations; so seek to mitigate them and show relevance relative to other methods Contact Information: Clifford A. Lipscomb, Ph.D. Director of Economic Research Greenfield Advisors, LLC 1870 The Exchange SE, Suite 100 Atlanta, GA 30339 USA E-mail: cliff@greenfieldadvisors.com Web: www.greenfieldadvisors.com Thank you