Estimating Home Equity Impacts from Rapid, Targeted Residential Demolition in Detroit, MI: Application of a Spatially-Dynamic Data System for Decision Support

A Report Produced by Dynamo Metrics, LLC1 July 2015

ABSTRACT

In an effort to further the goals of the Motor City Mapping Initiative, investments were made to create a data architecture for a spatially-dynamic decision support system in Detroit, Michigan. The data system that now exists is capable of tracking the time series dynamics of every one of the more than 384,000 parcels in Detroit between January 1st, 2011 and March 31st, 2015 on a quarterly basis (seventeen quarterly time steps for each parcel). To provide a rapidly produced and useful example of the analytic capabilities of the spatially-dynamic data system, it was used to estimate the effect of the in progress, rapidly deployed Hardest Hit Fund (HHF) demolition investment concentrated in selected areas of Detroit between April 1st, 2014 (Q2 2014) and March 31st, 2015 (Q1 2015). This study utilizes causal modeling that incorporates spatially- dynamic econometric methods in the context of a spatio-temporal treatment effects analysis to estimate the impact of the HHF Blight Elimination Program implementation on single-family home values in Detroit.

Using the year prior to HHF implementation (Q2 2013 – Q1 2014) as a control for the rapid and targeted HHF implementation (Q2 2014-Q1 2015), findings suggest that home equity increases of up to 13.8%2 exist for single-family homes that sold inside HHF demolition zones (HHF Zones) after the implementation of the HHF Blight Elimination Program. Further results suggest each demolition event within HHF Zones after policy implementation nets a 4.2% positive impact on the value of nearby homes, while single-family home counterparts outside the HHF Zones net a 2.1% positive impact on the value of nearby homes during the same time period from nearby demolition activity. Findings thus suggest that home equity protection hedges created by demolition activity are greater, and homes are more valuable overall, within HHF Zones after HHF implementation than elsewhere in the city. It can therefore reasonably be maintained that the HHF Blight Elimination Program is having a market-stabilizing effect on the neighborhoods it targets.

1 Report authors can be reached at [email protected] or www.dynamometrics.com. 2 The “up to” 13.8% suggests that more research is warranted to determine the portion of the 13.8% that is specifically caused by HHF demolition activity, and what portion, if any, can be attributed to the many other positive neighborhood activities, investments and associated effects that may have stemmed from targeted HHF investments. 1 Acknowledgements

The team at Dynamo Metrics would like to thank Rock Ventures, LLC, and The Skillman Foundation for helping make this project possible and for their dedication to the increased sophistication of data systems in Detroit. Also, thanks to all partners of the Motor City Mapping initiative and those departments in the city of Detroit that provided data for this study for their commitment to a rich data environment in Detroit. Thanks to Data Driven Detroit and LOVELAND Technologies, LLC for project support and help integrating into the Detroit data movement.

2 Table of Contents Acknowledgements ...... 2 Introduction ...... 4 Increasing the Sophistication of Detroit’s Data Ecosystem...... 4 Constructing a Spatially-Dynamic Data System in Detroit ...... 4 Identifying Useful Application of a Spatially-Dynamic Data System in Detroit ...... 4 Estimating Home Equity Impacts from Rapid, Targeted Residential Demolition in Detroit, MI ...... 6 Empirical Research Approach ...... 6 Data ...... 6 Empirical Model ...... 7 Empirical Findings ...... 9 Interpretation of Key Empirical Findings ...... 11 Policy Implications ...... 15 Study Limitations and Future Research ...... 16 References ...... 17 Appendices ...... 19 1. Cautionary Note on Summary Statistics ...... 20 2. Summary Statistics of Average Single-family Homes: Inside/Outside HHF Zones & Before/After HHF Implementation ...... 21 3. Summary Stats of Average Single-family Homes: Housing Submarkets Based on Block-Group Level Median Income ...... 22 4. Summary Statistics of Average Single-family Homes: Median Income Housing Submarkets Inside/Outside HHF Zones ...... 23 5. Summary Statistics of Average Homes: Inside/Outside HHF & Before/After HHF in High Income Submarket ...... 24 6. Summary Statistics of Average Homes: Inside/Outside HHF & Before/After HHF in Middle-High Income Submarket ...... 25 7. Summary Statistics of Average Homes: Inside/Outside HHF & Before/After HHF in Middle-Low Income Submarket ...... 26 8. Summary Statistics of Average Homes: Inside/Outside HHF & Before/After HHF in Low Income Submarket ...... 27 9. Map of Submarkets for this Analysis: 5-Year Census Estimates of Median Income at Block Group Level ...... 28 10. Map of Demolition Concentrations Before HHF Implementation (Q2 2013-Q1 2014) ...... 29 11. Map of Demolition Concentrations After HHF Implementation (Q2 2014-Q1 2015) ...... 30 12. Map of Average Single-Family Sale Price Concentrations Before HHF (Q2 2013-Q1 2014) ... 31 13. Map of Average Single-Family Sale Price Concentrations After HHF (Q2 2014 – Q1 2015) ... 32 14. Map of Unoccupied Tax Foreclosable Residential in 2015 Q1 ...... 33 15. Map of Vacant Lot Concentrations in 2015 Q1 ...... 34 16. Map of Average Sales Price Change After HHF Blight Elimination Program Implementation 35 17. Chow Test Results - HHF Regimes Diagnostics ...... 36

3 Introduction

Increasing the Sophistication of Detroit’s Data Ecosystem

To further the goals of the Motor City Mapping initiative, The Skillman Foundation and Rock Ventures, LLC, sought to develop data analytics capabilities for Detroit. Dynamo Metrics, LLC (Dynamo), was hired for two tasks. First, Dynamo was to construct a spatially-dynamic and analytics ready data system from available public and proprietary information. Second, Dynamo was to produce an analytics product derived from this data system to show the usefulness of a sustained commitment to data infrastructure.

From December 2014 through May 2015, Dynamo constructed the spatially-dynamic data system. During this time, it was decided that the Detroit policy community may benefit from an analytics product focused on the impact of the HHF Blight Elimination Program investments now underway. Dynamo explored both correlative and causal modeling approaches. In June 2015 Dynamo determined that it could construct a causal modeling approach, although the explanatory power of the causal model would be less robust than in normally-functioning markets. Time constraints limited Dynamo’s ability to exhaustively test and stress the modeling process, as would be expected in an academic environment. Those limitations notwithstanding, this report’s results provide insight, performance measurement, and decision support to Detroit’s leadership, and show examples of what can be quickly produced if Detroit makes a sustained commitment to data infrastructure. Constructing a Spatially-Dynamic Data System in Detroit

A time series oriented data system that tracks change in -level and neighborhood-level characteristics is critical to performing analysis on the dynamics of property values, home occupancy, , crime, blight, vacant lots, and other localized neighborhood health phenomena. Understanding these dynamics creates the ability to test performance and quantify return on investment (ROI) from various policies and programs. To this end, Dynamo created a data system in Detroit (Detroit Space-Time Analytics Data System™, or D-STADS™) that tracks property characteristics and their quarterly dynamics at the parcel level. D-STADS™ currently tracks from first-quarter of 2011 through the first-quarter of 2015. D-STADS™ allows for customizable property intervention analytics enabling quantification of positive and negative externalities associated with events such as nearby properties being sold, blighted properties being demolished, nearby properties being renovated, and beyond. The inherently spatial architecture of D-STADS™ further allows for the production of dynamic illustrations of unfolding events in static and online map-based visualization formats.

Identifying Useful Application of a Spatially-Dynamic Data System in Detroit

In 2010 U.S. Treasury’ Hardest Hit Fund (HHF) made $7.6 billion available for foreclosure prevention in 18 states, including Michigan. Michigan’s share was over $498 million. In 2013 Michigan was allowed to reprogram $100 million of this HHF money for demolition of blighted residential structures. Detroit was allocated $52.3 million of the $100 million. Since then Michigan has allocated an additional $75 million of HHF money to blight elimination, with $50 million of it committed to Detroit. U.S. Treasury allowed Detroit to spend HHF money on blight

4 elimination because previous research3 demonstrated through causal modeling that blight demolition preserves surrounding property values. That research also established strong correlation between blight elimination and reduction in mortgage foreclosure rates.

This study builds upon that previous research. This study further tests the theory that demolition of blighted residential structures in targeted neighborhoods creates positive property value impacts for nearby houses in those neighborhoods. The primary difference between previous blight elimination research and this research application is that demolition occurred in targeted zones4 and that demolition was rapidly performed. This circumstance and an operationalized D-STADS™ allow for concise space-time testing of the HHF Blight Elimination policy in terms of how rapid, concentrated demolition impacts residential property value. Such testing may reasonably lead to consideration of further targeted and rapid blight elimination investment in Detroit and similar communities nationwide.

3 See: Griswold, N., Calnin. B., Schramm, M., Anselin, L. and P. Boehnlein. 2014. Estimating the Effect of Demolishing Distressed Structures in Cleveland, OH, 2009-2013: Impacts on Real Estate Equity and Mortgage-Foreclosure. Publication of the Western Reserve Land Conservancy: Thriving Communities Institute. 4 Ibid. 5 Estimating Home Equity Impacts from Rapid, Targeted Residential Demolition in Detroit, MI

Empirical Research Approach

The delineation of the HHF Zones and the implementation of the HHF Blight Elimination Program in Detroit can be viewed as “natural experiments” or “treatments” in a quasi- experimental sense (for methodological background, see, e.g., the recent reviews in Angrist and Pischke, 2015, and Imbens and Rubin, 2015). Policymakers are interested in the effect of the rapid and targeted demolition experiment on specific outcome variables, such as residential house sales prices. A typical analysis of these effects is regression discontinuity (a change in the time trend for an outcome variable after the treatment is applied) or difference-in-differences (comparing the before-and-after difference in outcome between treated and untreated observations). The HHF Blight Elimination Program is more challenging, however, because the sales in question pertain to different properties, so there are no observations (sales) before the experiment (demolition) for a property that sold after the experiment. As a result, a traditional regression discontinuity or difference-in-differences analysis cannot be carried out. In this study’s regression analysis, this challenge is addressed by separating the sales observations into different “regimes” (within HHF Zones and outside of HHF Zones), and only then comparing the value attributed to key policy variables in a hedonic regression before and after the HHF implementation (the second quarter of 2014) (for details on spatial regimes regression, see Anselin and Rey, 2014). In other words, this study measures how selling inside/outside HHF Zones, and before/after HHF Blight Elimination Program implementation, impact house sale prices.

Given that targeted and rapid demolition went into full motion in April of 2014, the best data that D-STADS™ has during the policy covers one year (four quarterly time-steps): Q2 2014 - Q1 2015. A proper control group was constructed by extending the observation period back in time to Q2 2013 to create a two-year total time series for the study (Q2 2013 - Q1 2015). April 1st, 2014 is the before and after break of single-family sales observations given it is the estimated start-date of the HHF demolition program.

Data

An intensive process of data quality control and assessment was undertaken to make sure the data was consistent, as error free as possible5, and linked across each data source in a common format to achieve readiness for causal and other statistical modeling. Property status in terms of sales, occupancy, crime, and all other data used for analytics were identified for every property in every quarterly time-step in Detroit. Data from the following sources was incorporated:

 Detroit City Assessor (Property characteristics; Property ownership transfers)  Wayne County Treasurer (Tax foreclosure eligible properties)  Valassis Corporation (Postal vacancy)

5 Please contact the authors for explanation of specific data processing details and issues within the Detroit data system.

6  City of Detroit Police Department (Crime)  Detroit Open Data Portal (Parcel shapefile)  Data Driven Detroit (HHF Zone boundaries)  US Census Bureau (Tigerline Block Group shapefile and income data)

In addition, each property received a summary of the characteristics and conditions of all properties within 500 feet of it during each quarterly time step. With the dynamic change of the immediate environment surrounding each individual property accounted for over time, analysis which tracks the causalities and key trends of change can be studied. The result of this critical step in data processing is the Detroit Space-Time Analytics Data System™.

This study used D-STADS™ for the limited purpose of analyzing the impact of HHF demolitions carried out through the first quarter of 2015 as an exercise in its analytic capacity. If D-STADS™ is kept current and robust it can become a powerful tool for decision makers, researchers and practitioners alike. The future use of D-STADS™ is envisioned as the heart of an online user- friendly decision support tool with an online map-based visualization capability and interface that allows users to analyze and simulate customizable property interventions and understand their impact. Empirical Model

Empirical methods associated with the hedonic price function are based on the economic theory that goods are ultimately valued by way of their utility-bearing attributes (Lancaster, 1966; Rosen, 1974). While specific application of these methods may vary, contemporary literature provides strong evidence of the negative spillover effects on real estate values from nearby distressed properties (Simons, Quercia and Maric, 1998; Immergluck and Smith, 2006; Griswold, 2006; Griswold and Norris, 2007; Schuetz, et al., 2008; Mikelbank, 2008; Leonard and Murdoch, 2009; Harding, et al., 2009; Rogers and Winter, 2009; Lin, Rosenblatt and Yao, 2009; Kobie 2003; Rogers, 2010; Hartley, 2010; Campbell et al., 2011; Groves and Rogers, 2011; Whitaker and Fitzpatrick, 2013; Griswold and Calnin, et al. , 2014). The empirical models used in this study leverage D-STADS™ to build upon established model design and evidence. Here, the model is specifically designed to test the impact of distressed housing and vacant lots on housing values in the context of the implementation of the HHF Blight Elimination Program as a “natural experiment” or “treatment” in a quasi-experimental environment (Angrist and Pischke, 2015, and Imbens and Rubin, 2015).

The specific empirical model design for this research application is as follows:

푅 퐻퐻퐹 ln(푃푖) = 훽0 + 훽1(퐷푖 ) + 훽2(푃푖 ) + 훽3(푆푖) + 훽4(푆푀푖) + 훽5(푆푉푖) + 훽6(푆푇푖) + 푢푖

Where Pi , the natural log of Detroit’s single-family home housing prices between Q2 2013 – Q1 2015 - is determined by:

7 푅 1. Several vectors from several spatial variables, 훽1(퐷푖 ), measuring the aggregate count of multiple types of residential structures and vacant lots within 500 feet of a residential sale during the quarter in which it sold6; 퐻퐻퐹 2. Several vectors from several HHF policy variables, 훽2(푃푖 ), estimating the effect of before (Q2 2013-Q1 2014) and after (Q2 2014-Q1 2015) HHF implementation, as well as the effect of being inside and outside the HHF impact zones; 3. Several vectors of variables describing the physical attributes of sold houses, 훽3(푆푖); 4. Several vectors of indicator variables, 훽4(푆푀푖), identifying the Detroit income-based housing submarket a house sold in. Submarkets were identified at the block group level using a 5-year U.S. Census Bureau estimate of Median Household Income; 5. Several vectors of indicator variables, 훽5(푆푉푖), identifying the quarter a home sold in an attempt to control for the seasonal effects of homes sales; 6. Given Detroit’s distressed market, several vectors of indicator variables, 훽6(푆푇푖), account for the sale/ type of the house that sold, and;

7. The error term, ui , in all analyses accounts for the presence of heteroskedasticity of unknown form.

Three versions of the empirical model were run in this analysis to provide necessary insight into the effect of distressed structures, vacant lots and rapid, targeted HHF demolition intervention. First, a citywide version of the model was run which incorporates indicator variables identifying: 1) whether a single-family home sold inside or outside the targeted HHF Zones; and 2) whether a single-family home sold before or after the implementation of the HHF Blight Elimination Program. Second, a separate regression was run for only those sales that occurred within an HHF designated zone. Third, a separate regression was run for the sales occurring outside the designated HHF Zones. These two separate regressions can be thought of as “spatial regimes” and taken together in a “regimes regression” they allow an assessment of the extent to which the model coefficients differ between regimes, hence an assessment of the difference between the housing markets inside and outside HHF Zones. That said, they are presented separately to allow the contrast between sales behavior inside and outside HHF Zones. The regimes model requires that the HHF “inside/outside” variable be dropped given that it was the indicator variable used to split the citywide observations into the two respective regimes. Each model – citywide and regimes - possess distressed structure and vacant lot counts, as well as associated before/after HHF interaction variables that quantify the home value effect of turning blighted structures into vacant lots by way of nearby demolition events (500 feet in all model versions).

6 A unique space-time lag variable was designed for this analysis to capture highly localized changes in the housing market and neighborhood effects as they vary over time (see Anselin, 1988 and Anselin and Rey 2014 for technical details on the concept of a spatially lagged variable). It is constructed in the spirit of the “comparable sales” approach in . For each property sold in a given quarter, the average is computed of the sales prices of the sixth closest (nearest neighbor) properties that sold in the quarter before. In our analysis, we used a logarithmic transformation for the sales price variable and constructed the space-time lagged variable from those logs. Hence its coefficient can be interpreted as the percentage change in sales price resulting from a percentage change in the “comparable” sales in the previous period. Since this variable takes on a different value for each sale, and varies both across space and over time, it is deemed to be more adept at reflecting variations in local market (and by implication, neighborhood) conditions than a standard spatial fixed effect indicator variable. A spatial fixed effect variable would take on the value of 1 for all sales at all periods that happen in the same spatial area, typically a census tract or block group, and thus would not capture any changes over time (see Anselin and Arribas-Bel 2013, for a technical discussion).

8 Empirical Findings

Table 1 below provides results from the three versions of empirical model specified above. The total number of single-family home sales identified during the study time-period is 8,386, with 5,526 occurring inside the HHF Zone regime and 2,860 occurring outside the HHF Zone regime. The dependent variable in the regressions is based on the actual sales price of the 8,386 single- family home observations in Detroit over the study time period. Variables identified as explanatory in the regression of single-family house sales price are divided into six groupings: 1) the spatial variables provide counts of particular types of properties surrounding the observation of interest; 2) the policy variables are designed as indicator variables of whether a sales observation occurred inside/outside or before/after implementation of HHF as well as the interaction effect of some spatial variables after HHF implementation; 3) the physical variables explain the attributes of the single-family home that sold ; 4) the submarket variables are tiered into four levels of median income at the block group level; 5) the seasonal variables indicate which season a property sold in; and, 6) the sales type variables indicate which type of deed transfer occurred in the sale of the property. The regression analysis delivers “coefficients” associated with each variable we hypothesize to impact home sales price. These coefficient estimates portray the percentage effect that a given variable has on home sales price. Specifically, coefficients are read as percent impact on sales price from a marginal change in the variable – i.e. if bathrooms = “0.086*”, then one additional bathroom increases sales price by 8.6%. A coefficient can only be interpreted this way if the coefficient is statistically significant – i.e. “*”, “**” or “***” is placed next to the coefficient.

The following are some key summary statistics7 that provide a sense of the magnitude8 of the HHF Blight Elimination Program expenditures and the neighborhood context inside the HHF Zones that the dollars were spent in:

 A total of 3,739 demolitions occurred inside the HHF Zones between April 1st, 2014 (Q2 2014) and March 31st, 2015 (Q1 2015). o A total of 3,302 of the HHF Zone demolitions occurring after policy implementation were HHF funded demos (roughly 88.3%).

 The total number of occupied residential units inside the HHF Zones as of Q1 2015 is 112,763 occupied units. o 43.3% of total occupied residential homes (48,848) had an HHF demolition occur within 500 feet over the study time period.

 The average cost of an individual HHF demolition is $14,855 as of June 30th, 2015.

 The average price of “willing buyer/willing seller” sales of single-family homes inside the HHF Zones after the implementation of the HHF Blight Elimination programs is $26,336.

7 See Appendices 2-8 for detailed tables that provide all relevant observational summary statistics for inside/outside HHF Zones, before/after implementation of the HHF Blight Elimination Program, as well as some slicing up of the sub-marketing regime used for the study (can be viewed via map in Appendix 9). See Appendix 1 for write-up explaining that summary statistics presented in a hedonic modeling study can, if their function in the study is misunderstood, often appear to be at odds with the findings of the study. 8 Total demolitions inside HHF Zones between April 1st, 2014 – March 31st, 2015 and average HHF demolition cost provided by the Detroit Land Bank Authority on June 29th, 2015.

9

TABLE 1: Hedonic Regression of Single-family Homes Sales in Detroit, MI, Q2 2013 - Q1 2015 (8,386 Observations) Dependent Variable: Natural Log of Sales Price Citywide Inside HHF Outside HHF (8,386 Zones (5,526 Zones (2,860 Spatial Variables: Activity Within 500 Feet of Home Sold* Obs.) Obs.) Obs.)

# Residential Unoccupied Tax Foreclosable Structures -0.029*** -0.028*** -0.029*** # Residential Vacant Lots -0.005** -0.002 -0.008** # Occupied Residential Structures 0.003*** 0.003*** 0.003*** Violent Crime Rate -0.211 -0.666 0.251 Property Crime Rate 1.019** 0.416 2.021 # Properties Sold for >$25,000 0.097*** 0.103*** 0.089*** Ln of Average Sales Price of Nearest Six Neighbors That Sold in Previous Time Period 0.207*** 0.144*** 0.295*** Policy Variables: HHF Indicators Associated With Home Sold

Indicator of Sale Occurring Inside HHF Zones 0.145*** Indicator of Sale Occurring After HHF Implementation 0.11** 0.138*** 0.047 Additional Impact of Nearby Unoccupied Tax Foreclosable Structure After HHF -0.012* -0.014* -0.004 Additional Impact of Nearby Residential Vacant Lot After HHF Implementation -0.004 -0.006 -0.002 Physical Variables: Structural Attributes of Home Sold

Home Sold as Tax Foreclosable -0.582*** -0.548*** -0.639*** Square Footage of Home/100 0.035*** 0.037*** 0.029*** # Bathrooms 0.086* 0.052 0.137 # Fire Places 0.018 0.027 0.027 Brick Construction Indicator 0.207*** 0.212*** 0.223*** Square Footage of Porch/100 0.025 0.035* 0.016 Air Conditioning Indicator 0.057* 0.003 0.162** Age of Home When Sold -0.005*** -0.005*** -0.005** Submarket Variables: Block Group Level U.S. Census Median Income, 5-Year Est. Submarket 1: Middle Median Income Housing Market Pulled Indicator as Reference Submarket 2: Lowest Median Income Housing Market -0.104*** -0.111** -0.130* Submarket 3: Highest Median Income Housing Market 0.269*** 0.291*** 0.300 Submarket 4: Low Median Income Housing Market -0.072** -0.080** -0.088 Seasonal Variables: Indicators of the Quarter A Home Sold In Home Sold in First Quarter Pulled Indicator as Reference Home Sold in Second Quarter 0.039 0.020 0.081 Home Sold in Third Quarter 0.007 0.021 -0.012 Home Sold in Fourth Quarter -0.035 -0.017 -0.074 Sale Type Variables: Deed Transfer Type

Arms Length Sale Indicator 0.03 -0.006 0.091 Quit Claim Deed Indicator -0.818*** -0.911*** -0.640*** Warranty Deed Indicator 0.246*** 0.173** 0.381*** Land Contract Indicator 0.565*** 0.500*** 0.689*** (REO) Indicator -0.515*** -0.523*** -0.494*** Investment Property Indicator 0.453*** 0.409*** 0.551*** Hedonic Regression Constant 7.13*** 7.939*** 6.302*** Hedonic Regression R-Squared 0.324 0.296 0.337 Statistical Significance Key: *** = Less than 0.001; ** = Less Than 0.01; * = Less than 0.05.

10 The models performed marginally, with R-Squared values hovering around 0.30. This suggests that roughly 30% of the variation in the natural log of single-family home sales prices between Q2 2013 and Q1 2015 in Detroit are explained by this causal modeling exercise. The hedonic method used to specify the models in this study assume a rational market existing in perfect competition. Although great care was taken in identifying sales observations that include willing buyers and sellers of homes in Detroit, the market in Detroit is complex, is significantly impacted by economic hardship, speculation and tax foreclosure and changes drastically from block to block. These and many other factors provide a difficult space for market-based econometric modeling in Detroit. Nevertheless, even without its theoretical foundation based on market equilibrium, the model provides an adequate “empirical” representation of a range of determining factors in the house sales prices. A Chow test (Chow, 1960) was performed to test the difference in significance between the HHF and non-HHF regimes model versions. The global test showed high statistically significant difference between the two regimes, while independent variables were weakly different from one another – see Chow Test results in Appendix 16 for details.

With all weak-market and Detroit-specific modeling issues taken into account, the models in this study are quite stable and performed consistently in terms of the significance and magnitude of spatial, policy and other standard model variable coefficients. It is therefore reasonable to assume this model’s variables and associated coefficients would perform with stability and consistency in any future Detroit real estate market modeling. The relatively low R-Squared limits the value of the results in a purely predictive exercise. Nevertheless, the strong significance of the coefficient estimates for several of the important policy variables provides the necessary basis for insight and decision support to better understand the performance outcomes and neighborhood effect of the rapid, targeted HHF Blight Elimination Program in Detroit.

Interpretation of Key Empirical Findings

The spatial and policy variables are of significant interest in this exercise as they provide the necessary insight and decision support to estimate the performance and impact of the HHF Blight Elimination Program. The section below provides explanations of the value effect each key individual variable has on single-family home values in Detroit’s real estate market. The statements below are directly derived from Table 1, above.

Spatial Variable Interpretations

Residential Unoccupied Tax Foreclosable Structures Within 500 Feet of a Sale

 Citywide: Regression findings suggest with high statistical significance that an additional9 residential structure that is unoccupied and tax foreclosable within 500 feet of a single-family home in Detroit will negatively impact home value by 2.9%.  Inside HHF Zone Regime Regression findings suggest with high statistical significance that an additional residential structure that is unoccupied and tax foreclosable within

9 The word “additional” in economic research has a meaning slightly different that in normal English. When the word “additional” is used in this study, it can refer to the first event in a series, while in normal English “additional” never refers to the first event. So, when this study states “an additional residential structure” with certain characteristics, this can be referring to the first such residential structure, or the second, or the third, etc., in a series.

11 500 feet of a single-family home inside HHF demolition Zones will negatively impact home value by 2.8%.  Outside HHF Zone Regime: Regression findings suggest with high statistical significance that an additional residential structure that is unoccupied and tax foreclosable within 500 feet of a single-family home outside HHF Zones will negatively impact home value by 2.9%.

Residential Vacant Lots Within 500 Feet of a Sale

 Citywide: Regression findings suggest with high statistical significance that an additional residential vacant lot within 500 feet of a single-family home in Detroit will negatively impact home value by 0.5%.  Inside HHF Zone Regime: Regression findings suggest that a residential vacant lot within 500 feet of a single-family home inside HHF Zones will have no significant impact on housing value.  Outside HHF Zone Regime: Regression findings suggest with high statistical significance that an additional residential vacant lot within 500 feet of a single-family home outside HHF Zones will negatively impact home value by 0.8%.

Occupied Residential Structures Within 500 Feet of Sale

 Citywide: Regression findings suggest with high statistical significance that an additional occupied residential structure within 500 feet of a single-family home in Detroit will positively impact home value by 0.3%.  Inside HHF Zone Regime: Regression findings suggest with high statistical significance that an additional occupied residential structure within 500 feet of a single-family home inside HHF Zones will positively impact home value by 0.3%.  Outside HHF Zone Regime: Regression findings suggest with high statistical significance that an additional occupied residential structure within 500 feet of a single-family home outside HHF Zones will positively impact home value by 0.3%.

Properties that Sold For Greater Than $25,000 Within 500 Feet

 Citywide: Regression findings suggest with high statistical significance that an additional home that sells for greater than $25,000 within 500 feet of a single-family home in Detroit will increase single-family home sales by 9.7%.  Inside HHF Zone Regime: Regression findings suggest with high statistical significance that an additional home that sells for greater than $25,000 within 500 feet of a single- family home inside HHF Zones will increase single-family home sales by 10.3%.  Outside HHF Zone Regime: Regression findings suggest with high statistical significance that an additional home that sells for greater than $25,000 within 500 feet of a single- family home outside HHF Zones will increase single-family home sales by 8.9%.

12 Policy Variable Interpretations

Indicator of Sale Occurring Inside Targeted HHF Zones

 Citywide: Regression findings suggest with high statistical significance that a single- family home selling inside HHF Zones will sell for 14.5% more than those outside HHF Zones.

Indicator of Sale Occurring After HHF Demolition Program Implementation (Q2 2014 and later)

 Citywide: Regression findings suggest with high statistical significance that a single- family home sale in Detroit occurring after the implementation of the HHF Demolition Program sold with an 11% increase in value.  Inside HHF Zone Regime: Regression findings suggest with high statistical significance that a single-family home sale inside HHF Zones that occurs after the implementation of the HHF Demolition Program sold with a 13.8% increase in value.  Outside HHF Zone Regime: Regression findings suggest that a single-family home sale outside HHF Zones that occurred after the implementation of the HHF Demolition Program did not significantly impact value.

Interaction Variable of Unoccupied Tax Foreclosable Structures After HHF Demolition Program Implementation (Q2 2014 and later)

 Citywide: Regression findings suggest with statistical significance that an additional residential unoccupied tax-foreclosable structure within 500 feet of a single-family home sale has an additional negative impact of 1.2% on home value after the implementation of the HHF program.  Inside HHF Zone Regime: Regression findings suggest with statistical significance that an additional residential unoccupied tax foreclosable structure within 500 feet of a single- family home sale inside HHF Zones has an additional negative impact of 1.4% on home value after the implementation of the HHF Demolition Program.  Outside HHF Zone Regime: Regression findings suggest no statistically significant impact from an additional residential unoccupied tax foreclosable structure within 500 feet of a single-family home sale outside HHF Zones after the implementation of the HHF Demolition Program.

Interaction Variable of Residential Vacant Lots After HHF Demolition Program Implementation (Q2 2014 and later)

 Citywide: Regression findings suggest no statistically significant impact from an additional residential vacant lot within 500 feet of a single-family home sale after the implementation of the HHF program.  Inside HHF Zone Regime: Regression findings suggest with weak statistical significance that an additional residential vacant lot within 500 feet of a single-family home sale inside HHF Zones has an additional negative impact of 0.6% on home value after the implementation of the HHF Demolition Program.

13  Outside HHF Zone Regime: Regression findings suggest no statistically significant impact from an additional residential vacant lot within 500 feet of a single-family home sale outside HHF Zones after the implementation of the HHF Demolition Program

14 Policy Implications

 Citywide model estimates a 14.5% increase in single-family home sale prices inside the HHF Zones. This finding suggests that homes inside HHF Zone boundaries are selling for more than outside, which may indicate proper targeting of HHF demolition funds given that a program goal is to positively impact home equity.

 The Inside HHF Zone regime model strongly suggests that homes within HHF Zones sell for up to 13.8% more within HHF zones and after implementation of the HHF Blight Elimination Program. Meanwhile, single-family homes selling outside HHF Zones had no significant increase in value associated with the post-HHF implementation variable. These coefficient estimates suggest that home sales within HHF Zones after HHF Blight Elimination Program implementation are better off. This strength could reasonably be attributed to implementation of the HHF Blight Elimination Program. But further analysis is required for a complete understanding of this finding. Study of the summary statistics in Appendix 1-7 can provide greater insight on this question.

 The blighted structure interaction variable is significant in the HHF Zone regime model, suggesting an increased home value impact associated with demolition of nearby blight after the implementation of the HHF program. The aggregate negative impact from a blighted structure within 500 feet inside HHF Zones after HHF implementation is 4.2%, while an estimated negative value impact of only 2.8% exists before HHF implementation inside the HHF Zones. This finding suggests that rapid, targeted demolition of nearby blight causes remaining blight to be increasingly negative in its impact on nearby homes – a potential signal that finishing the job of full eradication of blight may be best for the stability of neighborhood home equity.

 Vacant lots have no significant negative impact on value within HHF Zones, suggesting a positive value impact of 4.2% is available for nearby homes from demolition inside HHF Zones. On the other hand, outside HHF Zones vacant lots have a negative effect on surrounding home value, so that the aggregate positive value impact that is estimated as available from demolition outside HHF Zones is only 2.1%, both before and after HHF implementation.

In short, the observations contained in the paragraphs above suggest that the greatest home value increases from demolition activity appear to be inside HHF Zones after the implementation of the HHF Blight Elimination Program.

15 Study Limitations and Future Research

Now that D-STADS™ exists, there are several potential improvements of the hedonic model for future research of Detroit’s housing market:

 Extending the scope of local property characteristics, in particular with respect to various accessibility measures (distance to “main activities”, access to transit, local amenities such as parks, distance to schools/quality of schools, walk score, etc.);  Examining the potential of a multilevel modeling approach, utilizing extensive neighborhood characteristics;  Exploring the sensitivity of the results to the “range of influence” of blighted locations (e.g., beyond 500ft);  Refining the definition of submarkets10 (e.g., using clustering techniques, machine learning) and assessing the sensitivity of the results to submarket “regimes”;  Further examining the sensitivity of the results to the specification of the space-time dynamics of the market, e.g., demolitions in more than one quarter vs. current quarter, the range of nearest neighbors (in the space-time lag);  Further examining the sensitivity of the results to different specifications for the spatial spillover effects (e.g., same-quarter spatial lag in addition to space-time lag, spatial error autocorrelation).

10 See Appendix 8 for reference to the submarket approach used in this modeling exercise.

16 References

Angrist, Joshua D. and Jorn-Steffen Pischke (2015). Mastering metrics, the path from cause to effect. Princeton, NJ, Princeton University Press.

Anselin, Luc and Sergio Rey (2014). Modern spatial econometrics in practice. Chicago, IL, GeoDa Press.

Anselin, Luc and Daniel Arribas-Bel (2013). Spatial fixed effects and spatial dependence in a single cross-section. Papers in Regional Science 92, 3-17.

Anselin, Luc (1988). Spatial econometrics, methods and models. Dordrecht, Kluwer.

Campbell, J.Y., Giglio, S., Pathak, P., 2011. Forced sales and house prices. American Economic Review 101, 2108–2132.

Chow, G. 1960. Tests of inequality between sets of coefficients in two linear regressions. Econometrica 28, 591-605.

Griswold, Nigel. 2006. "The Impacts of Tax-Foreclosed Properties and Land Bank Programs on Residential Housing Values in Flint, Michigan." Michigan State University Master’s Thesis.

Griswold, N. and Norris, P. 2007. Economic Impacts of Residential Property Abandonment and the Genesee County Land Bank in Flint, Michigan. Land Policy Institute.

Griswold, N., Calnin. B., Schramm, M., Anselin, L. and P. Boehnlein. 2014. Estimating the Effect of Demolishing Distressed Structures in Cleveland, OH, 2009-2013: Impacts on Real Estate Equity and Mortgage-Foreclosure. Publication of the Western Reserve Land Conservancy: Thriving Communities Institute.

Groves, J.R., Rogers, W.H., 2011. Effectiveness of RCA institutions to limit local externalities: using foreclosure data to test covenant effectiveness. Land Economics 87, 559–581.

Hartley, D., 2010. The effect of on nearby housing prices: supply or disamenity? Federal Reserve Bank of Cleveland, Working Paper, pp. 10–11.

Harding, J.P., Rosenblatt, E., Yao, V.W., 2009. The contagion effect of foreclosed properties. Journal of Urban Economics 66, 164–178.

Imbens, Guido W. and Donald B. Rubin (2015). Causal inference for statistics, social and biomedical sciences, an introduction. New York, NY, Cambridge University Press.

Immergluck D. and Geoff Smith. 2006. The External Costs of Foreclosure: The Impact of Single- family Mortgage-foreclosures on Property Values. Housing Policy Debate. Vol. 17. pp. 57 – 79.

Kobie, T.F. 2009. Residential Foreclosures’ Impact on Nearby Single-Family Residential

17 Properties: A New Approach to the Spatial and Temporal Dimensions. Cleveland State University Dissertation. 138 pp.

Lancaster, Kelvin J. 1966. A New Approach to Consumer Theory. Journal of Political Economy. Vol. 74. pp. 132-56.

Lin, Z., Rosenblatt, E., Yao, V.W., 2009. Spillover effects of foreclosures on neighborhood property values. Journal of Real Estate Finance and Economics 38, 387–407.

Leonard, T., Murdoch, J.C., 2009. The neighborhood effects of foreclosure. Journal of Geographic Systems 11, 317–332.

Mikelbank, B.A., 2008. Spatial analysis of the impact of vacant, abandoned and foreclosed properties. .

Rogers, W.H., Winter, W., 2009. The impact of foreclosures on neighboring housing sales. Journal of Real Estate Research 31, 455–479.

Rogers, W.H., 2010. Declining foreclosure neighborhood effects over time. Housing Policy Debate 20, 687–706.

Rosen, S. 1974. “Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition.” The Journal of Political Economy. Vol. 82, No. 1. pp. 34 -55.

Schuetz, J., Been, V., Ellen, I.G., 2008. Neighborhood effects of concentrated mortgage- foreclosures. Journal of Housing Economics 17, 306–319.

Simons, R., Quercia, R., and Maric, I. 1998. The Value Impact of New Residential Construction and Neighborhood Disinvestment on Residential Sales Price, Journal of Real Estate Research. Vol 15. pp. 147-61.

Whitaker, S., and T.J. Fitzpatrick. 2013. Deconstructing distressed-property spillovers: The effects of vacant, tax-delinquent, and foreclosed properties in housing submarkets. Journal of Housing Economics. Vol. 22, pp. 79-91.

18 Appendices

19 1. Cautionary Note on Summary Statistics The summary statistics in Appendices 2-8 provide several cuts of the 8,386 homes sales observations that were used in this hedonic modeling exercise. Given that the sales observations were cut into many unique spatial zones, some value typologies (i.e. high value sales or low value sales observations) may be over- or underrepresented in the summary statistics tables. For example, in Appendix 5 the average sales prices in the high income housing submarket outside the HHF Zones are much higher than those inside the HHF Zones. Given that the hedonic model findings suggest that homes within the HHF Zones after the implementation of the HHF Blight Elimination Program sell for up to 13.8% more than those outside the HHF Zones, these summary statistics may be confusing. The reason for this potential confusion is that the HHF Zones are residential areas with relatively consistent housing market types – i.e standard single family homes in consistent markets. So even the higher-end of these markets have sales only in the $50,000 range. On the other hand, outside the HHF Zones the high-end market includes many higher values single-family homes – i.e. hundreds of thousands of dollars. In short, the model suggesting a 13.8% higher sales value for homes selling in HHF Zones is consistent and correct, even though summary statistics may lead a reader to think otherwise because extreme high end sales do not occur within HHF Zones. Other, similar misunderstandings can arise.

20 2. Summary Statistics of Average Single-family Homes: Inside/Outside HHF Zones & Before/After HHF Implementation

Outside HHF Inside HHF Before After Outside Inside Citywide Change Difference Before After Before After HHF HHF HHF HHF Change Change HHF HHF HHF HHF Count of Single-family Home Sales 8,386 4,637 3,749 -19.2% 2,860 5,526 93.2% 1,628 1,232 -24.3% 3,009 2,517 -16.4% Average Sales Price $22,531 $20,670 $24,833 20.1% $19,637 $24,029 22.4% $18,028 $21,763 20.7% $22,099 $26,336 19.2% Average Sales Price of Recent Nearby Sales $13,869 $12,748 $15,393 $2,645 $11,522 $15,265 $3,743 $10,748 $12,631 $1,882 $13,980 $16,958 $2,978 Percent Sold in First Quarter 21.7% 23.4% 19.6% -3.8% 21.7% 21.6% -0.1% 22.4% 20.9% -1.4% 23.9% 18.9% -5.1% Percent Sold in Second Quarter 27.5% 30.2% 24.0% -6.2% 28.8% 26.7% -2.1% 31.9% 24.8% -7.1% 29.3% 23.7% -5.6% Percent Sold in Third Quarter 26.5% 25.0% 28.2% 3.2% 26.5% 26.5% 0.0% 24.6% 29.0% 4.3% 25.3% 27.9% 2.6% Percent Sold in Fourth Quarter 24.4% 21.4% 28.2% 6.8% 22.9% 25.2% 2.2% 21.1% 25.3% 4.2% 21.5% 29.6% 8.1% Average Square Footage 1,087 1,083 1,093 0.9% 1,061 1,101 3.8% 1,054 1,070 1.5% 1,099 1,104 0.4% Average Age of Home When Sold 74 74 75 1.4% 74 74 0.6% 73 75 1.9% 74 75 1.2% Percent Sold With Air Conditioning 14.8% 15.2% 14.3% -0.9% 13.7% 15.4% 1.6% 14.1% 13.3% -0.8% 15.9% 14.8% -1.1% Average Number of Bathrooms 1.19 1.18 1.19 0.9% 1.16 1.20 3.2% 1.16 1.17 0.7% 1.20 1.21 0.9% Percent Sold Made of Brick 68.5% 67.7% 69.5% 1.8% 63.8% 70.8% 7.0% 62.8% 65.2% 2.3% 70.3% 71.6% 1.3% Average Number of Fire Places 0.4 0.4 0.4 4.5% 0.3 0.4 23.5% 0.3 0.4 9.1% 0.4 0.4 2.0% Average Porch Square Footage 105 103 106 2.4% 102 106 3.7% 100 105 5.6% 105 106 0.8% Percent Sold as Tax Foreclosable 12.6% 11.3% 14.2% 2.9% 14.1% 11.8% -2.3% 14.7% 13.2% -1.5% 9.4% 14.7% 5.3% Percent Sold as Arms Length Sale 4.4% 3.9% 5.0% 1.1% 4.7% 4.3% -0.4% 3.4% 6.4% 3.0% 4.2% 4.4% 0.2% Percent Sold as Quit Claim Deed 1.4% 0.6% 2.3% 1.8% 1.2% 1.4% 0.2% 0.6% 2.1% 1.6% 0.6% 2.5% 1.9% Percent Sold as Warranty Deed 3.8% 3.0% 4.6% 1.6% 3.7% 3.8% 0.0% 3.1% 4.6% 1.6% 3.0% 4.6% 1.6% Percent Sold as Land Contract 3.8% 3.9% 3.7% -0.2% 3.9% 3.8% -0.1% 4.7% 2.8% -2.0% 3.5% 4.1% 0.7% Percent Sold as Real Estate Owned (REO) 16.1% 23.4% 7.1% -16.3% 17.4% 15.4% -2.0% 25.0% 7.5% -17.5% 22.5% 7.0% -15.6% Percent Sold as Investment Properties 16.1% 19.6% 11.7% -7.9% 16.6% 15.8% -0.8% 19.3% 13.0% -6.4% 19.7% 11.1% -8.6% Average Number Homes Sold >$25K within 500 Feet 0.3 0.3 0.3 -3.9% 0.2 0.3 61.3% 0.2 0.2 -22.5% 0.3 0.3 1.4% Average Number of Unoccupied Tax Foreclosable Properties within 500 Feet 6 6 5 -18.5% 6 5 -12.4% 6 6 -14.6% 6 5 -20.4% Average Number of Occupied Homes within 500 Feet 77 77 77 0.4% 72 79 9.8% 72 72 0.0% 79 79 0.2% Average Number of Tax Foreclosable Homes within 500 Feet 25 27 21 -24.1% 26 24 -9.2% 29 22 -22.9% 27 20 -24.6% Average Number of Vacant Residential Lots within 500 Feet 8 8 9 3.8% 11 7 -31.5% 11 11 -1.3% 7 8 9.5% Average Number of Violent Crimes within 500 Feet 0.9 1.0 0.9 -7.8% 0.9 0.9 4.2% 0.9 0.9 -0.8% 1.0 0.9 -11.2% Average Number of Property Crimes within 500 Feet 2.7 3.0 2.3 -23.3% 2.4 2.9 18.1% 2.7 2.1 -24.4% 3.2 2.5 -23.3%

21 3. Summary Stats of Average Single-family Homes: Housing Submarkets Based on Block-Group Level Median Income Middle High Middle Low High Income Low Income Income Income Submarket Submarket Submarket Submarket Count of Single-family Home Sales 300 2,057 3,731 2,298 Average Sales Price $52,488 $27,672 $20,004 $18,121 Average Sales Price of Recent Nearby Sales $29,272 $17,417 $12,920 $11,510 Percent Sold in First Quarter 22.3% 19.9% 22.7% 21.4% Percent Sold in Second Quarter 26.3% 26.7% 27.4% 28.4% Percent Sold in Third Quarter 20.7% 26.6% 27.2% 25.9% Percent Sold in Fourth Quarter 30.7% 26.7% 22.7% 24.3% Average Square Footage 1,507 1,192 1,040 1,016 Average Age of Home When Sold 76 73 74 77 Percent Sold With Air Conditioning 23.0% 19.1% 13.9% 11.4% Average Number of Bathrooms 1.53 1.27 1.15 1.13 Percent Sold Made of Brick 83.3% 84.0% 69.2% 51.4% Average Number of Fire Places 0.71 0.53 0.37 0.26 Average Porch Square Footage 120 106 101 106 Percent Sold as Tax Foreclosable 14.3% 11.0% 13.5% 12.3% Percent Sold as Arms Length Sale 8.0% 4.8% 3.6% 5.0% Percent Sold as Quit Claim Deed 2.0% 1.0% 1.3% 1.7% Percent Sold as Warranty Deed 3.0% 3.2% 4.1% 3.9% Percent Sold as Land Contract 4.3% 3.2% 4.2% 3.7% Percent Sold as Real Estate Owned (REO) 17.7% 15.8% 16.7% 15.2% Percent Sold as Investment Properties 11.3% 17.4% 17.4% 13.3% Average Number Homes Sold >$25K within 500 Feet 0.5 0.3 0.3 0.2 Average # of Unoccupied Tax Foreclosable Properties within 500 Feet 2 4 6 7 Average Number of Occupied Homes within 500 Feet 73 79 77 74 Average Number of Tax Foreclosable Homes within 500 Feet 15 21 26 27 Average Number of Vacant Residential Lots within 500 Feet 5 5 8 13 Average Number of Violent Crimes within 500 Feet 0.6 0.8 1.0 1.0 Average Number of Property Crimes within 500 Feet 2.3 2.8 2.8 2.6

22 4. Summary Statistics of Average Single-family Homes: Median Income Housing Submarkets Inside/Outside HHF Zones

High Income Submarket Middle High Income Submarket Middle Low Income Submarket Low Income Submarket Outside Outside Outside Outside Inside HHF Difference Inside HHF Difference Inside HHF Difference Inside HHF Difference HHF HHF HHF HHF Count of Single-family Home Sales 22 278 1163.6% 550 1,507 174.0% 1,432 2,299 60.5% 856 1,442 68.5% Average Sales Price $148,879 $44,860 -69.9% $26,147 $28,228 8.0% $17,667 $21,459 21.5% $15,428 $19,720 27.8% Percent Sold in First Quarter 22.7% 22.3% -0.4% 19.6% 20.0% 0.4% 22.8% 22.7% -0.1% 21.4% 21.4% 0.1% Percent Sold in Second Quarter 27.3% 26.3% -1.0% 26.9% 26.6% -0.3% 29.1% 26.3% -2.7% 29.7% 27.7% -2.0% Percent Sold in Third Quarter 13.6% 21.2% 7.6% 27.6% 26.3% -1.4% 26.8% 27.4% 0.6% 25.6% 26.1% 0.5% Percent Sold in Fourth Quarter 36.4% 30.2% -6.1% 25.8% 27.1% 1.3% 21.4% 23.5% 2.2% 23.4% 24.8% 1.5% Average Square Footage 2,615 1,419 -45.7% 1,152 1,207 4.8% 1,021 1,051 2.9% 1,029 1,008 -2.0% Average Age of Home When Sold 84 76 -9.6% 73 73 -0.8% 73 74 1.1% 76 77 1.5% Percent Sold With Air Conditioning 31.8% 22.3% -9.5% 16.2% 20.2% 4.0% 13.6% 14.0% 0.4% 11.9% 11.2% -0.8% Average Number of Bathrooms 2.77 1.43 -48.6% 1.24 1.28 2.9% 1.14 1.17 2.7% 1.12 1.13 1.3% Percent Sold Made of Brick 77.3% 83.8% 6.5% 74.2% 87.6% 13.4% 65.1% 71.8% 6.7% 54.8% 49.4% -5.4% Average Number of Fire Places 0.2 0.8 230.8% 0.5 0.6 14.9% 0.3 0.4 21.8% 0.3 0.3 -7.9% Average Porch Square Footage 144 118 -18.3% 107 106 -1.4% 100 102 2.3% 101 109 8.1% Percent Sold as Tax Foreclosable 13.6% 14.4% 0.8% 13.3% 10.2% -3.1% 14.5% 12.9% -1.6% 13.9% 11.4% -2.5% Percent Sold as Arms Length Sale 4.5% 8.3% 3.7% 5.8% 4.4% -1.4% 4.5% 3.1% -1.4% 4.4% 5.3% 0.8% Percent Sold as Quit Claim Deed 4.5% 1.8% -2.7% 0.7% 1.1% 0.3% 1.2% 1.4% 0.2% 1.5% 1.7% 0.2% Percent Sold as Warranty Deed 4.5% 2.9% -1.7% 2.7% 3.3% 0.6% 3.9% 4.2% 0.3% 4.1% 3.7% -0.3% Percent Sold as Land Contract 0.0% 4.7% 4.7% 2.4% 3.5% 1.1% 5.0% 3.7% -1.3% 3.2% 4.0% 0.9% Percent Sold as Real Estate Owned (REO) 13.6% 18.0% 4.3% 14.7% 16.3% 1.5% 17.9% 16.0% -2.0% 18.5% 13.2% -5.2% Percent Sold as Investment Properties 9.1% 11.5% 2.4% 16.2% 17.8% 1.6% 18.3% 16.9% -1.4% 14.3% 12.7% -1.6% Average Number Homes Sold >$25K within 500 Feet 0.2 0.6 214.6% 0.3 0.4 32.4% 0.2 0.3 39.5% 0.1 0.3 111.9% Average Number of Unoccupied Tax Foreclosable Properties within 500 Feet 4 2 -48.3% 4 4 3.9% 6 6 -6.1% 7 6 -12.9% Average Number of Occupied Homes within 500 Feet 38 76 102.4% 78 80 3.1% 74 79 7.1% 66 79 19.0% Average # of Tax Foreclosable Homes within 500 Feet 14 15 7.2% 21 21 -0.4% 27 25 -5.8% 28 26 -10.0% Average # of Vacant Residential Lots within 500 Feet 11 4 -63.9% 6 4 -31.9% 10 7 -26.6% 15 11 -24.5% Average Number of Violent Crimes within 500 Feet 0.1 0.7 616.2% 0.8 0.8 9.1% 1.0 1.0 -0.4% 0.9 1.1 17.8% Average Number of Property Crimes within 500 Feet 1.0 2.4 143.5% 2.6 2.9 9.1% 2.5 2.9 18.9% 2.3 2.9 25.2%

23 5. Summary Statistics of Average Homes: Inside/Outside HHF & Before/After HHF in High Income Submarket Outside HHF Inside HHF Before HHF After HHF Change Before HHF After HHF Change Count of Single-family Home Sales 13 9 -30.8% 148 130 -12.2% Average Sales Price $111,987 $202,167 80.5% $38,380 $52,237 36.1% Percent Sold in First Quarter 7.7% 44.4% 36.8% 22.3% 22.3% 0.0% Percent Sold in Second Quarter 30.8% 22.2% -8.5% 33.1% 18.5% -14.6% Percent Sold in Third Quarter 15.4% 11.1% -4.3% 20.3% 22.3% 2.0% Percent Sold in Fourth Quarter 46.2% 22.2% -23.9% 24.3% 36.9% 12.6% Average Square Footage 2,461 2,838 15.3% 1,441 1,395 -3.2% Average Age of Home When Sold 86 81 -5.0% 74 77 4.1% Percent Sold With Air Conditioning 23.1% 44.4% 21.4% 23.6% 20.8% -2.9% Average Number of Bathrooms 2.42 3.28 35.3% 1.47 1.38 -6.3% Percent Sold Made of Brick 84.6% 66.7% -17.9% 85.1% 82.3% -2.8% Average Number of Fire Places 0.4 0.0 -100.0% 0.7 0.8 10.6% Average Porch Square Footage 125 171 36.8% 115 121 5.1% Percent Sold as Tax Foreclosable 0.0% 33.3% 33.3% 6.8% 23.1% 16.3% Percent Sold as Arms Length Sale 7.7% 0.0% -7.7% 14.9% 0.8% -14.1% Percent Sold as Quit Claim Deed 0.0% 11.1% 11.1% 0.7% 3.1% 2.4% Percent Sold as Warranty Deed 0.0% 11.1% 11.1% 1.4% 4.6% 3.3% Percent Sold as Land Contract 0.0% 0.0% 0.0% 2.7% 6.9% 4.2% Percent Sold as Real Estate Owned (REO) 15.4% 11.1% -4.3% 24.3% 10.8% -13.6% Percent Sold as Investment Properties 15.4% 0.0% -15.4% 18.2% 3.8% -14.4% Average Number Homes Sold >$25K within 500 Feet 0.2 0.2 44.4% 0.5 0.6 27.5% Average # of Unoccupied Tax Foreclosable Properties within 500 Feet 5 3 -46.9% 2 2 10.5% Average Number of Occupied Homes within 500 Feet 40 34 -16.9% 76 76 -0.3% Average Number of Tax Foreclosable Homes within 500 Feet 17 9 -45.8% 16 13 -19.2% Average Number of Vacant Residential Lots within 500 Feet 12 9 -25.1% 3 5 48.7% Average Number of Violent Crimes within 500 Feet 0.2 0.0 -100.0% 0.6 0.7 31.5% Average Number of Property Crimes within 500 Feet 0.9 1.1 20.4% 2.6 2.3 -12.1%

24 6. Summary Statistics of Average Homes: Inside/Outside HHF & Before/After HHF in Middle-High Income Submarket Outside HHF Inside HHF Before HHF After HHF Change Before HHF After HHF Change Count of Single-family Home Sales 274 276 0.7% 808 699 -13.5% Average Sales Price $23,284 $28,988 24.5% $25,987 $30,819 18.6% Percent Sold in First Quarter 21.9% 17.4% -4.5% 22.0% 17.7% -4.3% Percent Sold in Second Quarter 28.1% 25.7% -2.4% 30.3% 22.3% -8.0% Percent Sold in Third Quarter 26.3% 29.0% 2.7% 24.9% 27.9% 3.0% Percent Sold in Fourth Quarter 23.7% 27.9% 4.2% 22.8% 32.0% 9.3% Average Square Footage 1,151 1,153 0.2% 1,191 1,226 3.0% Average Age of Home When Sold 72 75 4.2% 72 74 2.4% Percent Sold With Air Conditioning 15.7% 16.7% 1.0% 21.0% 19.2% -1.9% Average Number of Bathrooms 1.24 1.25 1.0% 1.25 1.31 4.9% Percent Sold Made of Brick 75.2% 73.2% -2.0% 87.5% 87.7% 0.2% Average Number of Fire Places 0.5 0.4 -12.1% 0.6 0.5 -1.7% Average Porch Square Footage 107 108 0.5% 107 105 -1.9% Percent Sold as Tax Foreclosable 12.8% 13.8% 1.0% 7.4% 13.4% 6.0% Percent Sold as Arms Length Sale 5.5% 6.2% 0.7% 5.2% 3.4% -1.8% Percent Sold as Quit Claim Deed 0.0% 1.4% 1.4% 0.2% 2.0% 1.8% Percent Sold as Warranty Deed 0.7% 4.7% 4.0% 2.2% 4.6% 2.4% Percent Sold as Land Contract 2.6% 2.2% -0.4% 3.6% 3.3% -0.3% Percent Sold as Real Estate Owned (REO) 23.7% 5.8% -17.9% 25.5% 5.6% -19.9% Percent Sold as Investment Properties 17.9% 14.5% -3.4% 22.9% 11.9% -11.0% Average Number Homes Sold >$25K within 500 Feet 0.3 0.3 -0.7% 0.4 0.3 -20.4% Average # of Unoccupied Tax Foreclosable Properties within 500 Feet 4 4 -15.4% 5 4 -13.0% Average Number of Occupied Homes within 500 Feet 79 76 -3.7% 80 79 -1.3% Average Number of Tax Foreclosable Homes within 500 Feet 24 18 -25.6% 24 18 -23.2% Average Number of Vacant Residential Lots within 500 Feet 6 7 11.9% 4 5 33.8% Average Number of Violent Crimes within 500 Feet 0.8 0.8 -4.0% 0.9 0.7 -18.1% Average Number of Property Crimes within 500 Feet 3.0 2.3 -21.3% 3.3 2.4 -26.5%

25 7. Summary Statistics of Average Homes: Inside/Outside HHF & Before/After HHF in Middle-Low Income Submarket Outside HHF Inside HHF Before HHF After HHF Change Before HHF After HHF Change Count of Single-family Home Sales 844 588 -30.3% 1,262 1,037 -17.8% Average Sales Price $16,842 $18,852 11.9% $19,931 $23,319 17.0% Percent Sold in First Quarter 23.2% 22.1% -1.1% 26.1% 18.5% -7.6% Percent Sold in Second Quarter 32.0% 24.8% -7.2% 27.8% 24.5% -3.3% Percent Sold in Third Quarter 25.1% 29.3% 4.1% 24.2% 31.3% 7.1% Percent Sold in Fourth Quarter 19.7% 23.8% 4.1% 21.8% 25.7% 3.9% Average Square Footage 1,015 1,031 1.6% 1,051 1,052 0.0% Average Age of Home When Sold 73 74 1.3% 74 74 0.6% Percent Sold With Air Conditioning 14.8% 11.9% -2.9% 13.5% 14.8% 1.3% Average Number of Bathrooms 1.14 1.13 -1.0% 1.17 1.16 -0.5% Percent Sold Made of Brick 63.0% 68.0% 5.0% 70.7% 73.1% 2.4% Average Number of Fire Places 0.3 0.4 21.0% 0.4 0.4 2.9% Average Porch Square Footage 96 105 8.6% 101 103 2.2% Percent Sold as Tax Foreclosable 16.1% 12.2% -3.9% 11.6% 14.5% 2.9% Percent Sold as Arms Length Sale 2.8% 6.8% 4.0% 2.4% 4.0% 1.6% Percent Sold as Quit Claim Deed 0.7% 1.9% 1.2% 0.7% 2.3% 1.6% Percent Sold as Warranty Deed 3.2% 4.9% 1.7% 3.0% 5.6% 2.6% Percent Sold as Land Contract 6.4% 2.9% -3.5% 3.4% 4.1% 0.6% Percent Sold as Real Estate Owned (REO) 24.6% 8.3% -16.3% 22.7% 7.8% -14.9% Percent Sold as Investment Properties 20.6% 15.0% -5.7% 20.4% 12.5% -7.9% Average Number Homes Sold >$25K within 500 Feet 0.3 0.2 -39.3% 0.3 0.3 4.0% Average # of Unoccupied Tax Foreclosable Properties within 500 Feet 6 6 -14.2% 6 5 -25.6% Average Number of Occupied Homes within 500 Feet 74 75 1.1% 79 80 1.4% Average Number of Tax Foreclosable Homes within 500 Feet 29 23 -23.1% 28 21 -25.4% Average Number of Vacant Residential Lots within 500 Feet 10 10 -4.7% 7 7 1.7% Average Number of Violent Crimes within 500 Feet 0.9 1.0 2.4% 1.0 0.9 -7.4% Average Number of Property Crimes within 500 Feet 2.7 2.1 -24.2% 3.3 2.5 -23.2%

26 8. Summary Statistics of Average Homes: Inside/Outside HHF & Before/After HHF in Low Income Submarket Outside HHF Inside HHF Before HHF After HHF Change Before HHF After HHF Change Count of Single-family Home Sales 497 359 -27.8% 791 651 -17.7% Average Sales Price $14,687 $16,454 12.0% $18,540 $21,155 14.1% Percent Sold in First Quarter 21.5% 21.2% -0.4% 22.6% 20.0% -2.7% Percent Sold in Second Quarter 33.8% 24.0% -9.8% 30.0% 24.9% -5.1% Percent Sold in Third Quarter 23.1% 29.0% 5.8% 28.2% 23.5% -4.7% Percent Sold in Fourth Quarter 21.5% 25.9% 4.4% 19.2% 31.6% 12.4% Average Square Footage 1,031 1,026 -0.5% 1,018 997 -2.0% Average Age of Home When Sold 75 77 1.7% 77 77 0.5% Percent Sold With Air Conditioning 11.7% 12.3% 0.6% 13.0% 8.9% -4.1% Average Number of Bathrooms 1.12 1.12 0.1% 1.14 1.13 -0.2% Percent Sold Made of Brick 55.1% 54.3% -0.8% 49.2% 49.6% 0.4% Average Number of Fire Places 0.3 0.3 5.1% 0.3 0.3 -1.0% Average Porch Square Footage 100 102 2.1% 109 109 0.2% Percent Sold as Tax Foreclosable 13.9% 13.9% 0.0% 8.6% 14.7% 6.1% Percent Sold as Arms Length Sale 3.2% 6.1% 2.9% 4.0% 6.8% 2.7% Percent Sold as Quit Claim Deed 0.6% 2.8% 2.2% 0.6% 3.1% 2.4% Percent Sold as Warranty Deed 4.2% 3.9% -0.3% 4.2% 3.2% -0.9% Percent Sold as Land Contract 3.2% 3.1% -0.2% 3.5% 4.6% 1.1% Percent Sold as Real Estate Owned (REO) 26.6% 7.2% -19.3% 19.0% 6.3% -12.7% Percent Sold as Investment Properties 18.1% 8.9% -9.2% 15.4% 9.4% -6.1% Average Number Homes Sold >$25K within 500 Feet 0.1 0.1 -9.7% 0.3 0.3 19.7% Average # of Unoccupied Tax Foreclosable Properties within 500 Feet 8 7 -9.3% 7 6 -18.4% Average Number of Occupied Homes within 500 Feet 66 66 -0.3% 79 79 0.2% Average Number of Tax Foreclosable Homes within 500 Feet 31 25 -18.1% 29 22 -24.2% Average Number of Vacant Residential Lots within 500 Feet 14 15 5.5% 11 12 8.4% Average Number of Violent Crimes within 500 Feet 0.9 0.9 -0.6% 1.2 1.0 -13.9% Average Number of Property Crimes within 500 Feet 2.6 1.8 -29.3% 3.2 2.5 -21.4%

27 9. Map of Submarkets for this Analysis: 5-Year Census Estimates of Median Income at Block Group Level

28 10. Map of Demolition Concentrations Before HHF Implementation (Q2 2013-Q1 2014)

29 11. Map of Demolition Concentrations After HHF Implementation (Q2 2014-Q1 2015)

30 12. Map of Average Single-Family Sale Price Concentrations Before HHF (Q2 2013-Q1 2014)

31 13. Map of Average Single-Family Sale Price Concentrations After HHF (Q2 2014 – Q1 2015)

32 14. Map of Unoccupied Tax Foreclosable Residential Properties in 2015 Q1

33 15. Map of Vacant Lot Concentrations in 2015 Q1

34 16. Map of Average Sales Price Change After HHF Blight Elimination Program Implementation

35 17. Chow Test Results - HHF Regimes Diagnostics

Degrees of Variable name Freedom Value Probability # Residential Unoccupied Tax Foreclosable Structures 1 0.014 0.9066 # Residential Vacant Lots 1 2.436 0.1186 # Occupied Residential Structures 1 0.137 0.7109 Violent Crime Rate 1 0.725 0.3946 Property Crime Rate 1 5.040 0.0248 # Properties Sold for >$25,000 1 0.278 0.5977 Average Sales Price of Neighbors That Sold in Previous Time Period 1 11.427 0.0007 Indicator of Sale Occurring After HHF Implementation 1 1.492 0.2219 Additional Impact of Nearby Unoccupied Tax Foreclosable Structure After HHF 1 0.809 0.3685 Additional Impact of Nearby Residential Vacant Lots After HHF Implementation 1 0.693 0.4053 Home Sold as Tax Foreclosable 1 1.397 0.2373 Square Footage of Home/100 1 0.646 0.4217 # Bathrooms 1 1.070 0.3010 # Fire Places 1 0.000 0.9910 Brick Construction Indicator 1 0.032 0.8570 Square Footage of Porch/100 1 0.364 0.5462 Air Conditioning Indicator 1 7.017 0.0081 Age of Home When Sold 1 0.015 0.9018 Submarket 2: Lowest Median Income Housing Market 1 0.077 0.7812 Submarket 3: Highest Median Income Housing Market 1 0.001 0.9800 Submarket 4: Low Median Income Housing Market 1 0.023 0.8792 Home Sold in Second Quarter 1 0.839 0.3597 Home Sold in Third Quarter 1 0.236 0.6271 Home Sold in Fourth Quarter 1 0.660 0.4165 Arms Length Sale Indicator 1 0.900 0.3427 Quit Claim Deed Indicator 1 2.808 0.0938 Warranty Deed Indicator 1 3.538 0.0600 Land Contract Indicator 1 4.476 0.0344 Real Estate Owned (REO) Indicator 1 0.315 0.5746 Investment Property Indicator 1 4.811 0.0283 Hedonic Regression Constant 1 12.800 0.0003 GLOBAL TEST 31 96.834 0.0000

36