Abstract

This paper estimates spillover effects from a spatially-targeted redevelopment program, the Real Property Investment Grant (RPIG), on single-family homes in the City of Richmond. Part of Virginia’s Enterprise Zone program, the RPIG subsidizes the rehabilitation, expansion, and new construction of commercial real estate in derelict industrial districts. Using data from the City of Richmond and the Virginia Department of Housing and Community Development, I find that the RPIG generates a small but statistically significant increase in the prices of single-family homes within one city block of RPIG-instigated investment.

Introduction

      The Virginia General Assembly moved to change the structure of Virginia’s Enterprise Zone program in 2005 from a series of tax breaks to two direct grants: the Job Creation Grant (JCG) and the Real Property Investment Grant (RPIG). The JCG subsidizes the creation of jobs by offering businesses up to $800 per year for every permanent, full-time position created that pays at least twice the federal minimum wage and offers health benefits. The RPIG subsidizes real property investments made inside of Richmond’s Enterprise Zones, which are distressed urban areas targeted by the state for economic development. The program offers up to $100,000 per building for qualifying investments of less than $5 million and up to $200,000 per building for qualifying investments of $5 million or more. In total, the RPIG has contributed to at least $1.25 billion of private investment in commercial real estate throughout Virginia.1 In Richmond alone, RPIGs worth $18 million have subsidized private investments totaling almost $500 million.

      In a study commissioned by the Virginia Department of Housing and Community Development, which administrates the Enterprise Zone program (EZ), the Center for Urban and Regional Analysis at VCU (CURA) finds that the EZ program benefits distressed urban areas through both business growth and job growth.2 Moreover, CURA reports that EZ designation improves residential property values up to a quarter of a mile outside of the EZ.3 However, CURA also discusses the uncertainty investors face when applying for the RPIG. Because the JCG takes precedence in the EZ program’s budget, the RPIG does not always receive full funding. Some years, the proration rate for the RPIG fell to as low as 43 percent. Although recipients of the RPIG report that the EZ program did not fundamentally change the feasibility of their operations, shortchanged businesses may refrain from further investment.4 The purpose of the following analysis therefore is twofold: to determine whether RPIG-instigated investment increases the price of single-family homes within a single city block and whether this effect is mitigated by shortfalls in RPIG funding.

Literature Review

      A property’s location plays a large role in determining its price. This is because its location determines both the presence and magnitude of a variety of non-market interactions between the purchaser and the property. These non-market interactions are largely generated by public goods such as access to public parks or local businesses, as well as public “bads” such as noise or air pollution. As discussed in Tiebout (1956), house hunters and businesses alike can therefore be thought of as “consumer-voters”: rational economic agents that express their preferences for the consumption of public goods by moving to the municipality that most closely approximates their utility-maximizing mix of public goods and taxes.5 And it is in this way that consumer-voters can also choose neighborhoods within municipalities to satisfy their utility-maximizing mix of public goods and property prices.

Figure 1: A map of RPIG projects and single-family homes in Richmond. Figure 1: Spatial distribution of RPIG projects and single-family homes in Richmond.

      Many studies estimate that spatially-targeted redevelopment programs improve property values in adjacent neighborhoods through the creation of public goods. While business owners value the economies of agglomeration that are generated by such programs, homeowners value access to a thriving business district. In this way, redevelopment programs generate non-market interactions between house hunters and homes. Freedman (2013), for instance, finds that enterprise zone designation in Texas generated a 10.7 percent increase in median home values inside enterprise zones.6 Busso et al. (2013) finds that the establishment of federal empowerment zones - a similar program administered across the United States at the federal level - led to an average 28 percent rise in the value of homes inside the empowerment zones.7 Moreover, Krupka et al. (2009) estimates that empowerment zone designation led to an appreciation of owner-occupied housing approximately 25 percent faster relative to what would have occurred without the program.8 However, these studies only consider a property’s relationship to the boundaries of a spatially-targeted redevelopment program rather than its relationship to specific instances of state-sponsored investment. This paper contributes to the existing literature therefore by estimating the proximate spillover effects of RPIGs on residential properties both inside and outside of Richmond’s enterprise zones.

Data

      The Virginia Department of Housing and Community Development provided the dataset for the Real Property Investment Grant. The data spans from 2005 to 2015 and only covers projects completed in City of Richmond. It contains information on the location of the property, the total investment that qualifies for RPIG funds, the requested amount of RPIG funds, and the actual amount of RPIG funds that the Virginia General Assembly disbursed. The data also includes information on the type of project the RPIG went towards — either rehabilitation, expansion, or new construction. This dataset contains information on a total of 336 RPIG projects.

Variable Mean Std. Dev. Min. Max
Total Investment (2015 $) 1,406,999 4,394,092 51,825 63,103,707
RPIG Requested 81,201 53,031 3,396 200,000
RPIG Disbursed 56,249 39,336 2,263 200,000
Difference 24,951 24,518 0 134,628
Type of Project (%):
Rehabilitation 0.86
Expansion 0.08
New Construction 0.06

Figure 2: Summary statistics for 336 RPIG projects.

      The dataset for the residential properties is from the Property Assessor’s Office for the City of Richmond. It contains information on the attributes of every property in Richmond: the square footage of both the building and the land, the year the building was constructed, both the neighborhood and the census tract in which the property is located, as well as the number of stories. It also includes information related to the most recent sale of the property, the condition of the building as observed in 2015, a description of the external covering of the property, as well as whether the purchaser of the property was an individual or a corporation. The total number of residential properties sold following the completion of an RPIG project within one block is 636.

Variable Mean Std. Dev. Min. Max.
Price (2015 $) 237,846 147,832 1,800 964,000
Finished Size (Sq Ft) 1,882 731 364 5,177
Plot Size 3,502 2,715 396 31,500
Year Built 1914 31 1830 2014
Condition (%):
Very Poor 0.02
Poor 0.03
Fair 0.09
Average 0.35
Good 0.38
Very Good 0.15
Exterior (%):
Vinyl 0.17
Brick 0.6
Other 0.23
Purchaser (%):
Corporation 0.2
Private Homeowner 0.8

Figure 3: Summary statistics for 636 single-family homes.

Methodology

      Ordinary Least Squares regression (OLS) selects the parameters of a linear function of a set of explanatory variables by minimizing the sum of the squares of the differences between the observed dependent variable and those predicted by the linear function. The OLS estimator is consistent when the regressors are exogenous and unbiased when the errors are both homoscedastic and serially uncorrelated. Standard hedonic considerations are used to motivate an OLS equation of the form:
\[ \begin{aligned} \ SalePrice_i = B_0 + B_1Requested_i + B_3Difference_i + B_4FinishedSize_i + B_5PlotSize_i+ \\ \ B_6YearBuilt_i+B_7Condition_i +B_8Exterior_i+ B_9Purchaser_i+ B_1YearSold_i + \epsilon_i\ \end{aligned} \] where i indexes each residential property. The dependent variable of interest is the sale price of the home because it reveals the purchaser’s preferences for both the internal and external attributes of the property. Included among the external attributes is the primary independent variable of interest: the aggregate sum of requested RPIG funds within 400 feet of the home at the time of the sale.9 The requested amount of RPIG funds is the most accurate measure of investment attributable to the RPIG program because the state approves the disbursement of RPIG funds the year following the actual investment. The proration rate of requested RPIG funds varies by year, from 43 to 100 percent. As pointed out by CURA, the difference between requested funds and disbursed funds may harm a business’s financial health if it is dependent upon reimbursement for additional investment. So a control variable for this difference is included in the model. Among the controls for the internal attributes of the home are the square footage of both the building and the plot of land it sits on, the material of the building’s exterior, and the condition of the building as graded by the City of Richmond in 2015. Furthermore, because corporations seeking to purchase rental properties are likely to be comparatively more skilled at negotiating a lower price, a control variable for whether or not the purchaser was a private homeowner is included. A control for the year a property was sold is also included to capture unobserved city-wide variations in the housing market. The natural log of several of these variables was taken to control for right-skewness: the sale amount, the requested amount of RPIG funds, the difference in the amount requested and the amount disbursed, and the square feet of both the building and the property.

Variable Mean Std. Dev. Minimum Maximum
Total Investment (2015 $) 981,363 1,308,991 51,825 6,665,767
RPIG Requested 107,598 99,725.87 3,396 759,644
RPIG Disbursed 71,363 63,132.95 2,263 477,140
Difference 36,235 42,461.97 0 282,504
Rehabilitation (# of Projects) 1.58 1.45 0 10
Expansion 0.11 0.35 0 3
Construction 0.05 0.23 0 2

Figure 4: Summary statistics for the aggregate sum of RPIG variables within 400 feet of 636 single-family homes.

      There were several controls removed due to issues of multicollinearity; that is, several variables could be linearly predicted by the others. This leads to biased estimates of the coefficients in an OLS estimation. Although an ideal model would control for the total qualifying amount of commercial real estate investment within 400 feet of a home, the amount of RPIG funds requested is a linear function of this amount. Multicollinearity arose for the same reason when including controls for the number and type of RPIG-instigated investments.

Results

      The first regression does not contain any spatial fixed effects. This was to test the relationship between the house’s price and its internal attributes. In this model, almost all of the OLS coefficient estimates are statistically significant except for the requested amount of RPIG funds. As expected, the model estimates that a house sells for slightly less if the purchaser is a corporation rather than a private homeowner. And the estimated coefficients demonstrate that houses with brick exteriors or other exteriors (which are primarily houses made of stone or wood) sell for more than houses with vinyl exteriors. Yet while the size of the building is positively correlated with its price, the size of the property is negatively correlated. Although this is at odds with what common sense would dictate, it points to the necessity of including spatial fixed effects. Because houses further from the city’s central business district tend to have larger yards, it is likely that their tendency to sell for less drives this effect. Furthermore, the estimated coefficients for the year a house was sold as compared to 2005 are all positive and statistically significant at at least the .05 level.

      The second regression contains spatial fixed effects by census tract, of which there are 30. As expected, the square footage of the building and the property are now both positively and significantly correlated with a property’s price. However, whether the purchaser was a corporation or a private individual is now statistically insignificant. Yet after including spatial fixed effects by census tract, the requested amount of RPIG funds is statistically significant at the .05 level. In terms of the parameters of the model, a one percent increase in the requested amount of RPIG funds generates a .091 percent increase in the price of a home.

     Regression Results

No Spatial Fixed Effects Census Tract Neighborhood
Intercept 8.094 *** 4.829 ** 6.6207 ***
log(Requested) 0.054 0.091 * 0.0707 .
(1.274) (2.121) (1.854)
log(Difference) -0.004 -0.006 -0.007 *
(-1.170) (-1.571) (-2.560)
log(Finished Size) 0.650 *** 0.4248 *** 0.3482 ***
(8.657) (5.108) (4.576)
log(Plot Size) -0.136 * 0.1629 * 0.1839 **
(-2.480) (2.570) (3.178)
Year Built -0.0007 0.0004 -0.0002
(-0.776 ) (0.531) (-0.285)
Condition: Very Poor -1.057 *** -0.8625 *** -0.6826 ***
(-6.234) (-5.876) (-4.917)
Condition: Poor -0.790 *** -0.5471 *** -0.3167 *
(-4.868) (-3.875) (-2.382)
Condition: Fair -0.430 *** -0.1721 * -0.0645
(-4.537) (-2.043) (-0.813)
Condition: Good 0.311 *** 0.2727 *** 0.2660 ***
(5.207) (5.027) (5.245)
Condition: Very Good 0.338 *** 0.3139 *** 0.2974 ***
(4.269) (4.501) (4.496)
Exterior: Brick 0.456 *** 0.1828 * 0.0671
(6.193) (2.531) (0.955)
Exterior: Other 0.154 . 0.1561 * 0.0622
(1.948) (2.219) (0.950)
Purchaser: Corporation -0.238 ** -0.0319 0.0127
(-3.258) (-0.358) (0.152)
R-Squared 0.456 0.636 0.678
Adjusted R-Squared 0.434 0.602 0.651

     Notes: Standard errors are shown in parentheses; Year fixed effects not shown above;
     Significance codes: .001 ‘***’ .01 ‘**’ .05 ‘*’ .1 ‘.’

This model is comparably better than the first beause its adjusted r-squared indicates that 60 percent of the variation in price is explained by variation in the model’s parameters, as compared to 43 percent for the first model. However, it is debatable whether a census tract is an appropriate spatial fixed effect because it may create arbitrary distinctions between neighborhoods that are contiguous in nature.

      The third and final regression contains spatial fixed effects by neighborhood as defined by the City of Richmond, of which there are 27. What is most interesting about this model is that the difference between the requested amount of RPIG funds and the disbursed amount is significant at the .05 level, while the requested amount of RPIG funds is now only significant at the .10 level. However, the negative sign on the coefficient of the former indicates that as this difference grows, the price of a home will decline. This is evidence in support of CURA’s hypothesis that failing to fully reimburse businesses for the requested amount of RPIG funds can be harmful to the local economy. In terms of the parameters of the model, a one percent increase in the requested amount of RPIG funds generates a .07 percent rise in the price of a home. Although this coefficient estimate is slightly smaller than in the previous model, the adjusted r-squared rises to 65 percent. This indicates that neighborhoods as defined by the City of Richmond provide the best spatial fixed effects.

Discussion

      The primary shortcoming of these models is their reliance upon a static measure of distance - 400 feet, or one city block. A more sophisticated spatial model would remedy this. Satre et al. (2011), for instance, employs nonparametric estimation to find that the decay rate of another one of Richmond’s redevelopment programs - the Neighborhoods in Bloom program - is approximately half every 1,000 feet further outside of the boundary of the targeted neighborhood.10 Another way to improve the models above would be to include controls for levels of direct investment, both in commercial and residential properties. This could be accomplished by obtaining a complete record of work permits in Richmond, which contain estimates of the total cost of construction associated with each project. Furthermore, the City of Richmond offers a variety of incentives to developers on top of the RPIG; Historic Tax Credits and Community Development Block Grants, for example. However, the majority of these incentives are also spatially-targeted, so spatial fixed effects likely capture much of this variance. The issue that remains is that of crowding out. A common critique of spatially-targeted redevelopment programs is that they only attract investment to one neighborhood at the expense of another. So, while the house prices within 400 feet of an RPIG-instigated investment may rise, house prices elsewhere in the city might fall as a result. Nevertheless, the above OLS coefficient estimates demonstrate that the Real Property Investment Grant does have a positive effect on the prices of immediately adjacent single-family homes. However, the third model also offers evidence that this effect is mitigated by shortfalls in RPIG funding. So, in order to maximize the positive spillover effects of the Real Property Investment Grant, its administrators should ensure that businesses are fully reimbursed for qualifying investments.

  1. Adhikari, Sarin, Mike MacKenzie and John Accordino. “A Review of the Virginia Enterprise Zone Program.” Center for Urban and Regional Analysis at VCU. 2016. (2)↩

  2. Ibid. (3)↩

  3. Ibid. (71)↩

  4. Ibid. (83)↩

  5. Tiebout, Charles. 1956. “A Pure Theory of Local Expenditures.” The Journal of Political Economy 64, pp. 416-424.↩

  6. Freedman, Matthew. 2013. “Targeted Business Incentives and Local Labor Markets.” Journal of Human Resources 48, pp. 311–344.↩

  7. Busso, Matias, Jesse Gregory and Patrick Kline. 2013. “Assessing the Incidence and Efficiency of a Prominent Place Based Policy.” American Economic Review 103, pp. 897–947.↩

  8. Krupka, Douglas and Douglas Noonan. 2009. “Empowerment Zones, neighborhood change and owner-occupied housing.” Regional Science and Urban Economics 39, pp. 386–396.↩

  9. 400 feet is roughly equivalent to a single city block in downtown Richmond.↩

  10. Sarte, Pierre-Daniel, Raymond Owens III and Esteban Rossi-Hansberg. 2011. “Housing Externalities.” The Journal of Political Economy 118, pp. 485-535.↩