The Economic and Social Benefits of The Wilmington Land Bank Conservancy

Two complementary regression models are used in this study to capture both the direct effect of WNCLB activity on sale prices (the Price Model) and the indirect effect through vacancy reduction (the Vacancy Reduction Model) and its influence on nearby property values.

• The Price Model measures the effect of WNCLB revitalization efforts on nearby property values.
• The Vacancy Reduction Model measures the effect to which WNCLB interventions have had to reduce overall vacancy by neighborhood and estimates those resulting indirect impacts on nearby property values.

A regression model was used to separate the effect of land bank remediation from other factors by controlling or underlying vacancy conditions, structural housing characteristics. Neighborhood fixed effects also help to measure the effects specific to the WNCLB’s efforts. The coefficient generated from each model helps tell whether the effect is statistically significant and estimates a specific dollar amount in change of nearby property values.

• Treatment Effect: WNCLB treatment of blighted properties increases nearby property values in the surrounding neighborhood. From the model results, every 10 treatments in a block group increase total property value by 7%. Collectively, WNCLB’s treatment on vacant properties increased total property value across the city by $158 million.
• Vacancy Reduction Effect: Reducing neighborhood vacancy further increases surrounding property values. The model results indicate that every 10 treatments in a block group reduce expected vacancy by 2.16 units and increases total property value by 4%. Together, vacancy reduction effects of WNCLB treated properties increased citywide total property value by another $81 million.

With the two effects combined, WNCLB’s treatments on blighted properties over the years increased total property value in the City by $239 million, generating $600,000 local property tax revenue annually to support public services and the school district.

The model draws the method in Penn IUR’s study, The price effects of greening vacant lots: How neighborhood attributes matter, which estimates how blighted parcels concerted in Philadelphia by their Land Bank into maintained green spaces affects nearby property values. Building on this approach, through Penn IUR’s expertise, the study adapts and refines the methodology to fit the smaller sample size and local geography in Wilmington and better assess the impacts of the WNCLB’s activities.

Data

• WNCLB Disposition Data (2017-2025): This study defines parcels as treated once they arrive at disposition to calculate annual cumulative intervention counts.
• New Castle County Sales Data (2013–2024): the County sales records are filtered to Wilmington and market transactions (excluding non-arm’s-length, multi-deed, and same-year repeat sales.)
• Vacancy Data (2016 and 2025): Wilmington vacancy registration data for 2016 and 2025 was used as a control in the Price Model and in the Vacancy Reduction Model.
• Census Data (2017): At the block group level ACS 5-year data was used to control for sociodemographic metrics as well as land-use data to represent land-use patterns.

The analysis uses two complementary regression models to estimate both the direct effect of WNCLB activity on sale prices (Price Model) and the indirect effect through vacancy reduction (Vacancy Reduction Model) and its associated impact on nearby property values.

• The Price Model estimates how WNCLB revitalization affects sale prices by regressing the log of inflation-adjusted sales prices on cumulative WNCLB treatments in the property’s census block group (2017 through the sale year). It controls for vacancy, housing characteristics, and block group sociodemographic and land-use measures to control for neighborhood fixed effect.
• The Vacancy Reduction model measures the effect of WNCLB interventions on the number of vacant parcels by neighborhood. It regresses the change in the number of vacant properties in each census block group between 2016 and 2025 on the cumulative number of WNCLB treatments as of 2025. This model captures the vacancy-reduction pathway and is combined with the Price Model to estimate total increment value effects.

Since its launch in 2017, the Wilmington Neighborhood Conservancy Land Bank (WNCLB) has been revitalizing communities by rehabilitating and returning vacant properties at little to no cost to qualified individuals, promoting homeownership, economic development, and neighborhood renewal.

Econsult Solutions, Inc. (ESI), under advisement from Penn IUR, conducted a statistical analysis to examine whether and to what extent the cumulative impact of WNCLB’s interventions has reduced vacancy and increased property values. Within the citywide results, the analysis found a positive and significant impact of cumulative WNCLB interventions on the value of nearby properties.

The vacancy model outlined in Section 4 of this appendix also indicates that interventions completed by the WNCLB to treat vacant properties correspond with a reduction in nearby housing vacancy rates at an approximate ratio of one fewer vacant property for every five WNCLB interventions conducted within a block group. Because vacancy itself depresses housing prices in the model results, the vacancy reductions associated with WNCLB interventions generate an additional price premium beyond the direct treatment effect.

At the citywide level, the direct treatment effect contributes roughly $158 million in incremental property value, but the associated reduction in vacancy adds an additional $80 million. The total of both effects yields a total increase of $239 million in residential property value across properties in the City.

The remainder of this report is structured as follows:

• Section 2 discusses the methodological framework used to estimate the effects of WNCLB treatments.
• Section 3 describes the datasets that support the analysis.
• Section 4 discusses the empirical results of the models, as well as the variables and details employed.

2.1 Referencing Penn IUR Framework

Penn IUR’s research on The price effects of greening vacant lots: How neighborhood attributes matter1 shows that converting blighted parcels into maintained green spaces raises nearby property values, a mechanism analogous to WNCLB interventions. Their framework uses a two-step repeated-sales design suited to very large datasets: a repeated-sales hedonic model produces a quality-adjusted price index, and a subsequent treatment-effect regression—weighted by propensity scores—isolates the impact of greening from non-random program placement.

The model used in this approach requires extensive repeat-sales links and large samples (Philadelphia’s dataset includes roughly 84,000 usable repeat-sales observations over ten years). Considering that Wilmington’s market has only about 6,000 single-sale transactions and fewer than 1,200 pairs of repeat-sales across a comparable period, the method was refined under the guidance of Penn IUR to create a similar modeling approach more applicable to the geography and data of Wilmington sales and land bank data.

2.2 ESI Adopted Methodology

This methodology implements a simplified but robust approach to constructing regression models to evaluate the effect WNCLB interventions have on surrounding property values. Two complementary models are estimated including a Price Model and a Vacancy Reduction Model. Together, they capture both the direct price effect of WNCLB activity and the indirect effect resulting from vacancy reduction.

Price Model

The Price Model measures the effect of WNCLB property revitalization on property sale prices. This model estimates the dollar impact on property sales associated with WNCLB's efforts. The dependent variable is the natural log of inflation-adjusted sales prices for the period 2013–2024 (n=6,096). The primary regressor of interest is the count of cumulative WNCLB treatments, defined as the total number of WNCLB interventions that had occurred in a property’s census block group from the start of WNCLB activity since 2017 up to the year of that sale. The intervention measure uses historical data of WNCLB’s inventory and assumes that neighborhood effects become most apparent after multiple properties have been removed from blighted conditions after disposition.

Vacancy levels (2025 inventory) are included as a core control to distinguish the effect of WNCLB activity from underlying neighborhood distress. Additional controls follow the Penn IUR hedonic framework, incorporating structural housing characteristics and neighborhood fixed effects. This analysis added census tract fixed effects to capture time-invariant neighborhood conditions; tract serves as a neutral geographic proxy for neighborhood context. This study tested replacing tract fixed effects with detailed block group sociodemographic controls such as homeownership rate, share of Black householders, median income, educational attainment, vacant land share, and residential land share. Tract fixed effects are excluded from this version to avoid multicollinearity with block group measures.

Vacancy Reduction Model

The Vacancy Reduction Model measures the effect of WNCLB interventions on the number of vacant parcels by neighborhood. ESI estimates a model of vacancy change to test whether WNCLB interventions are associated with declines in neighborhood vacancy levels. The dependent variable is the change in the number of vacant properties between 2016 and 2025 for each block group. The key regressor is the cumulative number of WNCLB treatments as of 2025.

This second model captures the vacancy reduction pathway, which indirectly contributes to property value increases and is combined with the Price Model to estimate total incremental value effects.

The analysis utilizes several data sets including WNCLB disposition data, Wilmington residential property sales data, Wilmington vacant property data, and census ACS estimates data.

WNCLB Disposition Data2

WNCLB disposition data was used to construct the cumulative treatment measure for each block group and year. After geocoding the disposition records and retaining those with valid location information, 405 treated parcels with disposition dates remained from 2017 to 2025. These records are joined to census block groups by year in order to compile the cumulative count of WNCLB interventions in each block group up to each sale year. This cumulative treatment variable is then merged with the residential sales data by block group and sale year and serves as the primary independent variable in the analysis.

Figure 3.1: Cumulative Acquisition and Disposition, 2017-2025

Source: WNCLB (2025)

Most properties move through the organization quickly. The largest group of 116 properties were disposed of within one year of acquisition, while 77 sold in the same year they were acquired (Figure 3.2). Together, these account for nearly half of all dispositions. 63 properties required four to six years, with rare cases extending to 7–8 years.

Figure 3.2: Frequency of Number of Years Between Acquisition and Disposition, 2017-2025*

Source: WNCLB (2025)

*Note: 2025 data reflect a partial year through September, not a full year of operation.

Figure 3.3: Number of Properties in Inventory by Acreage, 2025*

Note: Inventory is as of September, 2025. 

Source: WNCLB (2025)

As of September 2025, WNCLB holds over 200 properties in its inventory. The portfolio consists of small urban lots typical of city residential parcels. The most common sizes are 0.02 acres (60 lots) and 0.03 acres (64 lots), representing over 60% of inventory. Larger properties were less frequent with just 4 properties that were 0.31 acres or larger, reflecting a focus by the WNCLB mostly on standard infill redevelopment (Figure 3.3).

Residential Property Sales Data

Residential property sales data from 2013 through 2024 was obtained from New Castle County. The raw file includes 9,345 observations for The City of Wilmington and the file contains transaction price, sale date, and address. Each sale is joined to the Wilmington City parcel data by address in order to link each transaction to the detailed property attributes, including building area, lot size, structure type, age, bedroom and bathroom counts, and other key housing characteristics necessary for hedonic modeling.

Several filters are applied to sales data to ensure that only valid, market-reflective transactions remain. Sales below $10,000 are removed to exclude non-arm’s length transfers as well as any records flagged as a multi-deed transaction since these are reported as a single price for multiple properties because they do not provide parcel-level valuations. In addition, for parcels with multiple sales within the same calendar year, only the earliest sale transaction is retained to avoid duplicated price information. After these steps, the final analytic sample contains 6,096 residential sales suitable for estimating price effects.

Vacant Property Data

The inventory of vacant properties from The City of Wilmington was used to calculate the number within each census block group. Each sale is then assigned the vacancy count of its corresponding block group to control for underlying neighborhood conditions. The 2016 vacant properties inventory is also used to create a comparison with the 2025 inventory to compute the change in vacancy over time. This vacancy difference is used separately in the vacancy model that evaluates whether WNCLB interventions contributed to reducing neighborhood vacancy.

Census Data

Census block group level data from the ACS 2017 five-year estimates provide the sociodemographic controls used in the analysis, including Black householder share, median household income, homeownership rate, and housing vacancy rate. This set of ACS estimates were selected to control for neighborhood conditions before the start of WNCLB activity. New Castle County parcel-level data are used to compute the share of vacant land and the share of residential land within each block group.

The results demonstrate a positive and statistically significant relationship between Land Bank interventions and nearby property values. The price model with census tract fixed effects performs strongest, outperforming models without geographic fixed effects and those using only block group characteristics. The price model with a census tract as a fixed effect is therefore used as the primary model for interpretation and valuation.

In the price model the coefficient of cumulative WNCLB treatments is 0.007 (p < 0.05), indicating that each intervention is linked to a 0.7% increase in nearby property values after accounting for structural, locational, and vacancy characteristics. At this rate, ten treatments within a block group correspond to an estimated 7% increase in property values. Among block groups with at least one treatment, the median and mean treatment counts are 6 and 8, respectively, suggesting that many neighborhoods experienced meaningful levels of intervention.

In the same model, number of vacant properties show a negative and statistically significant relationship with nearby property values. The vacancy coefficient is –0.017 (p < 0.01), meaning that removing one vacant property is associated with a 1.7% increase in property prices.

A separate vacancy reduction model examines whether WNCLB interventions contribute to reducing vacant properties over time. The coefficient on cumulative treatments is –0.216 (p < 0.10), implying that approximately five treatments correspond to one fewer vacant property between 2016 and 2025. While modest in magnitude, this pattern indicates that WNCLB efforts play a role not only in improving nearby property values but also in helping stabilize neighborhoods by slowing or reversing vacancy growth.

Taken together, the two models demonstrate a compounding effect. WNCLB treatments are associated with higher property values directly, and they also contribute to reducing vacancy, which independently raises prices. The interaction of these two mechanisms produces an additional price premium: multiplying the vacancy coefficient (–0.017) by the vacancy reduction coefficient (–0.216) yields an incremental effect of roughly 0.004. For a block group with ten treatments, this translates to an additional price increase of about 4% beyond the direct treatment effect.

Figure 4.1 Model 1 Results - Property Value Price Model

Figure 4.2 Model 2 Results: Vacancy Reduction Model

4.1 Illustrative Example from Results

To demonstrate how treatment impacts are translated into monetary gains, consider the block group shown in Figure 4.3 as an example where the average home value in this block group is $90,900, and the block group contains 475 housing units3, yielding an estimated aggregate residential value of $43.2 million. By 2025, this block group had received 10 WNCLB treatments.

The price model estimates a 0.7% increase per treatment, attributing to a 7% price premium to the surrounding properties for 10 WNCLB treatments. Applied to the block group’s housing stock, this produces an incremental property value of $3.1 million.

The vacancy reduction model implies that 10 treatments reduce the expected vacancy count by 2.16 units. Feeding this vacancy reduction into the vacancy coefficient from price model generates an additional 4% price premium, translating to $1.6 million of incremental property value. Combined, the two premiums generate an estimated $4.6 million increase in property value.

Figure 4.3 Example - Estimated Incremental Property Value Gains of One Block Group (100030022001)

4.2 Citywide Incremental Property Value

Using the same calculation framework applied in the block group example, the regression coefficients are scaled across all residential properties in Wilmington. Aggregated median housing values are combined with the estimated treatment premium and the vacancy reduction premium to produce citywide totals4. WNCLB interventions are associated with a substantial uplift in residential property value. The direct treatment effect contributes roughly $158 million, and the vacancy reduction adds another $80 million. Together, these yield an estimated citywide gain of about $239 million in incremental residential property value for block groups that received WNCLB interventions as of 2025.

Figure 4.4 Citywide Estimated Incremental Property Value Gains

4.3 Discussion and Limitations

Vacancy is controlled using the 2025 inventory only. A multi-year vacancy panel would provide a stronger basis for identifying dynamic neighborhood change. In this framework, vacancy functions as a static proxy for underlying distress rather than a time-varying neighborhood condition.

Interpretation of the treatment coefficient in the price model should be read as the effect of cumulative WNCLB interventions while holding all other variables constant including the vacant property variable. The separate vacancy reduction model, although intentionally simple, provides the empirical link between interventions and subsequent changes in vacancy. Together, these two relationships form the basis for the compound effect: interventions raise prices directly and indirectly through vacancy reduction.

The simplicity of the vacancy model also implies limitations. With only 89 observations (many of which have zero treatments) the model captures an association rather than a causal mechanism. The limited degrees of freedom constrain the inclusion of additional controls, and any such controls would likely raise endogeneity concerns. For example, if block group income were found to be negatively related to vacancy, one could not reasonably assume income is independent of treatment. Nonetheless, the model supplies the necessary marginal response of how vacancy changes with additional treatments to translate vacancy reduction into price effects using the structural estimate from the main price model.