Did the Community Renewal Tax Incentives Pirate Businesses From Other Places?

Abstract

An early concern regarding place-based economic developments was that they might encourage business relocation into target neighborhoods at the expense of other places. To address this concern, when the U.S. Congress passed the Empowerment Zone/Enterprise Community (EZ/EC) program, it included an “antipirating” provision prohibiting use of grants for business relocation. Later iterations of the EZs and Renewal Communities (RCs) received tax incentives only. The RC program did not include an antipirating provision. Did businesses relocate? This study compares business moves within 1,000 feet inside a given RC/EZ with moves within 1,000 feet outside the RC/EZ before and after the intervention. Data are from the National Establishment Time Series (NETS) Database for California and Tennessee. Moves into some RC/EZs increased but so did moves out, leading to no statistically significant net change in numbers of firms. There were no obvious differences between RCs and EZs. The article concludes with policy recommendations.

Keywords

economic development tax incentives urban policy empowerment zones business relocation

To stimulate businesses and create jobs in high-poverty areas of inner cities, in 1993, the U.S. Congress authorized grants and tax incentives for the Empowerment Zone/Enterprise Community (EZ/EC) initiative (Omnibus Budget Reconciliation Act of 1993). Some scholars have expressed concern about this place-based approach (Levine, 1999; Quigley, 1994), given that it was inefficient to subsidize a business move from one location where it was profitable to a location where it would not be profitable without the subsidy. Furthermore, Quigley (1994) also argues that it would be more efficient to invest in human capital so that people could get better jobs, or use housing vouchers with increased enforcement of antidiscrimination laws to allow people to live in neighborhoods close to jobs. In contrast, regarding state and local government incentives, Bartik (1991) argues, generally, that competition for relocation of firms might help redistribute jobs to high unemployment areas where they are needed most, which might in turn reduce aggregate unemployment and inflationary pressures.

Congress anticipated concerns regarding businesses’ relocation from one jurisdiction to another based on the experience of the states. The statute addressed the concern of relocation with an “antipirating” provision that prohibited local government from using EZ/EC resources to move a business from one jurisdiction to another using grants or loans. Specifically, the antipirating provision prohibits state and local government applicants from including in the EZ strategic plan:

any action to assist any establishment¹ in relocating from one area outside the nominated area to the nominated area, except that assistance for the expansion of an existing business entity through the establishment of a new branch, affiliate, or subsidiary . . . (Taxpayer Relief Act of 1997, Pub. L. 105–34, title II, §226(b)(1), 111 Stat. 824, 1997)

In other words, local governments could use grants, loan funds, or bond proceeds to assist the opening of a branch, subsidiary, or affiliate of a business if the business did not shift jobs or close the original business at a later date.²

Congress passed the Community Renewal Act of 2000, which authorized eight additional Round III EZs and 40 Renewal Communities (RCs). These 48 designated areas received only tax incentives. The RC did not include an antipirating provision, possibly because Congress assumed that tax credits alone would not entice a business to relocate (U.S. Department of Housing and Urban Development, 2001, 2011). The Round I EC program ended in 2004, the Renewal Community program ended in 2009, and all EZs were extended to the end of 2013 (U.S. Department of Housing and Urban Development, 2013).

Closing the RC/EZ programs did not end the policy debate about using targeted programs to aid high-poverty places. In the first term of the Obama Administration (2011), the budget proposed a new Growth Zone program in 20 cities to replace the RC/EZ programs. Republican challenger Herman Cain suggested including an EZ-inspired program called the “999 plan” to reform the tax code (Schultz, 2011). In its second term, the Obama Administration renamed the proposed program Promise Zones (PZ; Office of the Press Secretary, 2013). Clearly, elected officials still found place-based programs compelling. Were businesses more likely to relocate to an RC/EZ than to neighboring areas? Were they less likely to move out? Was the antipirating prohibition a success? Was it even necessary? This study will first review the literature that presents evidence of the efficacy of the RC and EZ programs to assess the rationale for the PZ program as well as derive an identification strategy to estimate the impact on establishment relocations. We do not attempt to investigate violations of the statute, but rather to examine whether concerns raised by the statute were grounded and if having such a provision influenced establishment behavior. The primary contribution of this study is to be the first to directly measure the relocation of specific businesses into latter Round II and Round III EZs and RCs from outside using address-level establishment move data from the NETS database (Walls, 2008).

Previous Literature on Empowerment Zones and Establishments

Place-based federal urban policy has evolved from the Urban Renewal program in the 1950s, which involved massive redevelopment projects in low-income neighborhoods, to the Model Cities program in the 1960s, which gave grants to community-based organizations to plan to stabilize rather than replace existing neighborhoods. The 1970s and 1980s ushered in an era of new federalism, so urban policy focused mostly on investing in people in poverty through housing vouchers and income transfers, such as the Earned Income Tax Credit (Halpern, 1995; O’Connor, 1999; Quigley, 2000). In contrast to the broader policy shift to new federalism, place-based business investment found a home in the Carter administration’s Urban Development Action Grants, which gave grants to local governments to subsidize large commercial, industrial, or housing projects (Rich, 1989). During the 1980s, U.S. Congressman and the U.S. Department of Housing and Urban Development (HUD) Secretary Jack Kemp, encouraged states to experiment with U.K.-inspired “Enterprise Zones,” place-based tax incentives targeted to distressed areas (Lavin & Whysall, 2004). However, local governments’ ability to attract businesses was limited, leading to a practice referred to as “shoot anything that flies, claim anything that lands” (Rubin, 1988). It was in this context of conspicuous relocation that Congress passed the EZ/EC initiative in 1993, along with its terse antipirating provision. Scholars noted similarities between the community-based model cities approach and the tax incentive-oriented Enterprise Zone (Hetzel, 1994; Wallace, 2003, 2004).

Program Selection and Benefits

These programs had multiple eligibility and selection criteria. First, HUD required that Round I EZ/EC census tracts have high unemployment and at least 20% poverty, but there were exceptions made in Rounds II and III (U.S. Department of Housing and Urban Development, 1998). Second, the EZ/EC program allowed up to three noncontiguous areas. For example, the Los Angeles EZ had one part in the San Fernando Valley, one downtown, and one in south Los Angeles. For the RC program, the tracts had to average about 9% unemployment. It required that all tracts be contiguous but had a more generous size limit and did not provide exceptions to the 20% poverty requirement (U.S. Department of Housing and Urban Development, 2001). The primary criterion for selection among eligible applicants for EZ/ECs was the quality of the strategic plan submitted (U.S. Department of Housing and Urban Development, 1995). The EZ/EC regulations enshrined four strategic planning principles: (a) a strategic vision for change, (b) community-based partnerships, (c) economic opportunity, and (d) sustainable community development. The RC selection was based on severity of poverty and unemployment and did not require extensive community-based planning.

The resources available for RCs and EZs were substantial. Round I EZs received $100 million in Social Services Block Grants (SSBG) and tax incentives. Although using these grants to assist in business relocation was prohibited, nothing would preclude a business from relocating to take advantage of tax incentives. In particular, there was a wage credit worth 20% of qualifying wages for up to $3,000 for businesses that hired employees who lived and worked inside the EZ. Also included were increased Section 179 capital expensing, a partial exclusion on capital gains, and tax-exempt facility bonds for businesses that hired EZ residents and met other criteria. HUD also designated 65 ECs that received $3 million SSBG grants and inclusion of resident youth in the Work Opportunity Tax Credit (Taxpayer Relief Act of 1997, Pub. L. 105-34, title II, §226(b) (1), 111 Stat. 824, 1997; U.S. Department of Housing and Urban Development, 1995, 2011). The RC tax incentives included, among other incentives, an employment credit worth up to $1,500 when hiring those who lived and worked in the RC, a partial exclusion of capital gains taxes, and an accelerated depreciation schedule for businesses that hired RC residents; they also increased 179 deductions (U.S. Department of Housing and Urban Development, 2001, 2011). It is reasonable to expect some labor-intensive establishments to consider relocation to take advantage of the wage credit.

Further analysis of these programs is important because they are still active policy proposals. If authorized by Congress, President Obama’s PZ would include two tax incentives (Office of Management and Budget, 2011). The first would be a wage credit similar to the EZ wage credit, where businesses that hired employees who both lived and worked in the EZ received a credit up to 20% of the first $15,000 in qualifying wages. In the proposed PZ, however, businesses that hire PZ residents who work outside of the EZ would receive a credit equal to 10%. In this scenario, the employee lives in the PZ, but the work could be anywhere. The second PZ incentive is an accelerated depreciation deduction similar to the RC program in which establishments that placed buildings in service would be able to deduct an additional 100% of the adjusted basis of the property in the first year. In lieu of an antipirating provision, the PZ wage credit proposal relaxes the constraint that a business need be in a PZ. However, the scenario that allows a taxpayer-funded relocation to take advantage of an accelerated depreciation is still on the table.

Summary of Program Evaluations

The earliest studies assessed the first two EZ/EC strategic principles by looking at the role of governance (Chaskin & Peters, 1997; Gittell, Newman, & Pierre-Louis, 2001; Hebert, Vidal, Mills, James, & Gruenstein, 2001), community capacity, civic opportunity, and community participation (Gittell & Newman, 1998; Gittell, Newman, Bockmeyer, & Lindsay, 1998; Nathan & Wright, 1997; Oakley & Tsao, 2006). Because the primary eligibility criteria included tract-level 1990 poverty and unemployment numbers, studies investigated the effectiveness of the program to reduce poverty, reduce unemployment, increase jobs, or increase employment (Busso, Gregory, & Kline, 2010; Busso & Kline, 2008; Ham, Swenson, Imrohoroglu, & Song, 2011; Hanson, 2009; Hebert et al., 2001; Kolko & Neumark, 2010; Lee, 2005; Oakley & Tsao, 2006; Rich & Stoker, 2010; Smith, 2015; U.S. Government Accountability Office, 2006a). Other studies more relevant to this article investigated changes related to Principle (c) above (economic opportunity), such as business openings, survival, or industry mix (Hanson & Rohlin, 2011b; Hebert et al., 2001; Rich & Stoker, 2010; Robinson, 2005; Smith, 2011, 2014; U.S. Government Accountability Office, 2006a). Since housing was an expressed feature of Principle (d) above (sustainable community development), several studies estimated the impact of the EZ on real estate outcomes (Busso et al., 2010; Busso & Kline, 2008; Hanson, 2009; Krupka & Noonan, 2009a, 2009b; Rich & Stoker, 2010; Smith, 2007).

Although the EZ/EC literature does not include a relocation study per se, it does include studies about new establishments and employment outcomes whose design informs this study. The remainder of this literature review focuses on these studies. They each used different identification strategies that informed the choices used for this study: (a) variations on pre–post design with adjacent comparison groups, (b) propensity score matching, (c) instrumental variables, and (d) random growth curve. Each method and choice of variables allowed for different insights, advantages, and limitations.

For those that used matching, most reported both pooled and unpooled estimates. Using a pre–post design with a comparison group composed of adjacent high-poverty tracts in the same city, the HUD Interim Impact Assessment was the first major published impact study (Hebert et al., 2001). It finds that job growth increased in all Round I EZs except Chicago and Philadelphia. Minority and resident-owned businesses increased and reported hiring other residents. Another set of official impact studies by the U.S. Government Accountability Office (2004, 2006a, 2006b, 2010) combines a pre–post design with propensity score matching using proprietary business data from Caliper Corporation on jobs and establishments. Control census tracts were from within a 5-mile buffer around each of the eight Round I EZs. These studies conclude that results for the Round I EZs are mixed. They find that the Cleveland, New York, and Camden zones had statistically significant increases in numbers of businesses, while Atlanta, Baltimore, Chicago, Detroit, and Los Angeles saw decreases. Job growth increased statistically significantly in Baltimore, Detroit, Upper Manhattan, Philadelphia, and Camden, but decreased in Atlanta, Chicago, Cleveland, Los Angeles, Philadelphia, and the Bronx (U.S. Government Accountability Office, 2006a). Because of a small sample size and the inability to differentiate concurrent federal programs, the Government Accountability Office did not believe that it could make a causal claim (U.S. Government Accountability Office, 2006a).

In the first academic impact study, Oakley and Tsao (2006) use propensity score matching to identify controls within each city and conclude from a pooled regression that the program had no impact. Whereas Oakley and Tsao (2006) note improvements in specific zones (e.g., lower poverty in Chicago and Detroit), they caution that these improvements may have been the result of citywide spillover effects into the EZ. Rich and Stoker (2010) argue that pooling estimates across EZs is inappropriate because each EZ really employed a different policy treatment as each city customized its strategic plan. After matching tracts within each Round I city except Camden, they conclude that jobs and business investment, measured by the 3-year moving average of business loans in constant dollars, improved in all EZs except Atlanta but was not statistically significant because of the small sample size. They argue that the hypothesis test is less important than the effect sizes for policy purposes because local government is interested in seeing changes and is less concerned about whether the change happened by chance. In the first article to assess later rounds of EZs, Smith (2015) uses propensity score matching for tracts within the same state and finds statistically significant positive impacts on jobs and business openings, with a small reduction in business closings. However, results vary by zone. For example, the Chattanooga and San Francisco RCs did not see a statistically significant increase in jobs. Although the Knoxville EZ saw a statistically significant increase in jobs, it also saw a statistically significant decrease in business openings, suggesting a shift to more labor-intensive establishments. The Santa Ana EZ, Los Angeles RC, San Diego RC, and San Francisco RC each saw statistically significant reductions in business closures; however, Smith (2015) does not distinguish openings and closures from relocations.

Another set of articles uses an instrumental variable approach combined with difference-in-differences to estimate treatment effects of the Round I EZs (Hanson, 2009, 2011; Hanson & Rohlin, 2011a, 2011b). The comparison groups in these articles include both within-city tracts and EC tracts to control for selection bias and within-city spillover effects. Indeed, Hanson and Rohlin (2013) show that the Round I EZs have negative spillovers in tracts adjacent to the EZ and to other high-poverty tracts in the same city. In other words, the comparison areas appear to be harmed by the RC/EZ. The two-stage least squares instrumental variable adjusts the estimate of the treatment effect using a selection model based on Wallace (2003, 2004). Hanson and Rohlin (2011b) find that the share of industry employment shifted by 0.16 percentage points in the retail sector and by 30 percentage points in the service sector, but were offset by losses in finance, insurance, and real estate. The implication is that industries that are labor intensive may crowd out other industries by using the wage credit to outbid on real estate. Hanson and Rohlin (2011a) find that the first rounds attracted 20 new establishments, with most of the growth in retail and service sectors. However, they were not able to distinguish between new establishments and relocations in their data. This design estimates a local average treatment effect (LATE), which is the difference between two levels of treatment, with an instrument to adjust for political selection bias at the level of designation and an additional adjustment for exogenous variables.

The literature on utilization of tax incentives has been challenged because precise location information is not tracked by the Internal Revenue Service, but it has made national-level estimates available (Hanson, 2011; U.S. Government Accountability Office, 1999, 2004). The implication is that since the 1990s, the use of the RC/EZ wage credit went from approximately $10 million in 1997 to $180 million in 2008 for 1,040 filers alone (U.S. Department of Housing and Urban Development, 2010). These filers are individual taxpayers who have business income as sole proprietors and S-corporations and do not include corporate tax filers, so the actual utilization of the tax incentives is much larger. One of the reasons for the increasing utilization of wage credits is that consulting firms have integrated tax incentives into payroll, human resources applicant recruiting, and business location decision models (ADP, 2010; First Advantage, 2010). Furthermore, Hanson (2011) estimates that about one fourth of employees inside an RC or EZ are used by businesses to qualify for the wage credit. In summary, there is evidence of tax incentive utilization and mixed evidence of improvements in business and jobs outcomes. Thus, it is a reasonable hypothesis that some of the businesses-taking incentives moved to do so and that this may be observed in aggregate. However, to date, no study has directly investigated the movement of establishments into and outside of the RC/EZ.

There is a broader literature on the dynamics of establishment location that includes work on the location and relocation choices of startup establishments. Theoretically, establishment (re)location is a function of variable costs related to location (neoclassical theory); internal characteristics of a firm such as age and size (behavioral theory); and the social and institutional context of the establishment such as supply chain relationships and local taxes (institutional theory; Arauzo-Carod, Liviano-Solis, & Manjón-Antolín, 2010; Brouwer, Mariotti, & van Ommeren, 2004). Studies that use the establishment as the unit of analysis use discrete choice models and account for internal and external factors, and studies that use the location as the unit of analysis use count data models, such as Poisson or negative binomial regression. Of particular relevance to the EZ/EC tax incentives, Arauzo-Carod et al. (2010) find statistically significant negative elasticity in all 5 of the 28 reviewed location studies that had significant elasticity in taxes. According to Arauzo-Carod et al. (2010), the literature partially supports each of the three theories; however, most studies examine start-up location and not relocation because of data limitations on the precise location of establishments. It was their view that, all things equal, new establishments would have similar preferences as relocations. Therefore, with findings about business openings from the U.S. Government Accountability Office (2006a), Hanson and Rohlin (2011a), and Smith (2015) provide some insight into relocations.

One contribution of this study is that it overcomes the data limitations related to using census data. Outcome data are annual, address-level data that give a level of temporal and spatial precision. The NETS database contains both the origin address of establishments before a move as well as the destination location of an establishment. Did establishments increase moving from outside the RC/EZ to inside the RC/EZ? Did the RC/EZ reduce the relocation of establishments outside of the RC/EZ? Did branches and affiliates behave differently than headquarters and stand-alone establishments? This research is important for policy makers because the EZ incentives have been extended and new ones proposed.

Method

Selection Bias

Selection bias is a problem for intervention science in general and EZs in particular. The first selection problem involves the legislative drafting of the selection criteria that was arguably made with knowledge of conditions in anticipated neighborhoods (Wallace, 2004). Consequently, cities with large populations in poverty and which had high unemployment had greater or even certain probability of winning an award (Wallace, 2003, 2004). The second selection problem was the decision and capability of a local government to apply, and for which level of award (EZ or EC). Studies that allow comparison tracts from nonapplicant local governments can address this selection problem, and this study does not; thus, the parameter estimates are not generalizable. The third selection problem involves the local government’s choice, if any, of what combination of census tracts to include in the application. The concern about selection bias here is that the city could have used local knowledge to pick neighborhoods that they knew were going to improve without the EZ/EC program. If so, the effect of the program would appear more successful than it really was. Studies like this one that select comparison areas from within the same local government jurisdiction partially address this selection problem. The final step in the process is the selection by HUD, performed by teams of civil servants and political appointees. For the EZ/EC program, after screening for eligibility criteria the final winners were chosen based on a subjective assessment of the strategic plans. Studies that compare outcomes among winners with eligible applicants account for selection by HUD. Of the four selection problems, this study adjusts for the third kind (within-city) discussed in this paragraph by using time series data and a comparison group along the border to adjust estimates. This LATE parameter is robust to the selection of specific neighborhoods at the local level, conditional on having already been designated. The primary threat to internal validity would be unobserved time-varying confounding variables.

Identification Strategy: Adjusted Interrupted Time Series

This study uses an interrupted time series design (Morgan & Winship, 2007) along the border of the RC/EZ. Specifically, it uses a version of that design for community development interventions called an adjusted interrupted time series design (Galster, Temkin, Walker, & Sawyer, 2004) to estimate a local average intent to treat on the treated parameter (LATE). In other words, the region 1,000 feet inside of the RC/EZ is compared with the region 1,000 feet outside of the RC/EZ both before and during the policy treatment period. This buffer distance is the one used by Kolko and Neumark (2010). It is only possible to estimate the intent-to-treat (Rosenbaum, 2002) because there is no observed information on which businesses took the tax incentives and which did not.

For the purposes of comparing business moves, the best way to find a comparable area is to use one across the street. In the case of businesses, this implicitly controls for unobserved qualities of a particular location and local policy effects. Ham et al. (2011) use a similar identification strategy in their article on EZs by selecting tracts along the border. The strength of using adjacent areas, rather than low-income areas matched to the treatment group by a propensity score, lies in the assumption that for some small bandwidth of distance, any differences in the areas would be trivial. However, this LATE intent-to-treat estimate may be biased because benefits of the RC/EZ may have spillover effects in terms of jobs and business establishment openings and closings. As noted in the literature (Hanson & Rohlin, 2011a, 2013), spillovers may be positive when, for example, an improvement or increased retail traffic on one side of the street make the other side of the street attractive. In the context of the RC/EZ wage credit, although qualifying employees must live inside the nominated area and perform substantially all of the work inside of the nominated area, the establishment and the establishment’s owner may live anywhere. Furthermore, the Work Opportunity Tax Credit may be applied to a youth who lives in the RC/EZ, regardless of where the youth works. This introduces a positive spillover effect in the program and biases the estimate of the treatment effect downward. The spillovers may be negative, however, if public and private investment is targeted to the RC/EZ area at the expense of investment in other places. The literature suggests that spillovers on business openings in EZs are negative, but the impact on relocations has not been directly estimated (Hanson & Rohlin, 2011a, 2013). Under negative spillover effects, the estimate of the treatment effect of the RC/EZ is biased upward when adjacent areas are used as a counterfactual.

Data and Sample

Outcome data are from NETS, which has address-level data on establishments in the United States from 1990 through 2000. NETS assembles January snapshots of Dun & Bradstreet data r to hold seasonal variation constant. It has an advantage over Census ES-202 data sets such as the Longitudinal Employer Household Database, which only reports changes at the firm level (Neumark, Zhang, & Wall, 2005). Business codes not eligible for tax incentives were removed (e.g., nonprofit organizations, government, golf courses, suntan places, massage parlors, gambling establishments, and hot-tub facilities).

Although coverage of NETS is national, the licensing fee is cost prohibitive. To maximize the sample size while minimizing the cost of data purchase, California and Tennessee were selected. Marcuse (1997) argues that EZs would be a better fit for immigrant enclaves than for high-poverty African American neighborhoods. California data were already available at my institution, and California RC/EZ/ECs are among those with the highest proportion of foreign born. The Tennessee data were purchased because that state had the lowest proportion of foreign born among the RC/EZs and still had three designated areas with high-poverty African American neighborhoods. This provided some geographic and demographic variation, while remaining more cost effective than other state data packages. Results cannot be generalized and are not representative of other cities.

This study’s sample includes primarily Round II, Round III, and Renewal Community winners. The Round II and III areas are not as distressed on average as the Round I EZ/ECs, and the grant benefits were lower in value. The anticipated program effect sizes should be smaller than a sample of Round I EZs because the Round II and III program inputs are less valuable. Program impacts may also be influenced by the different level of distress in each RC/EZ and the relative experience of public and private sector officials working with the tax incentives.

The sample contains the following EZs and RCs: (a) Fresno, California, Round III, containing industrial and warehousing districts along Interstate 5; (b) Los Angeles, California, EZ and RC are treated as one area because they are adjacent, containing the area near the Burbank airport, Chinatown, the Fashion District (e.g., American Apparel), Koreatown, the Old Bank District downtown, and industrial areas in south Los Angeles; (c) San Diego, California, RC, a region including areas east of the airport through Gaslight Village downtown and east through neighborhoods; (d) San Francisco, California, RC containing blocks along Market Street from downtown to a diverse commercial and residential district called the Mission; and (e) Santa Ana, California, Round II, containing portions of downtown, an industrial park, a regional shopping center, and the Santa Ana Zoo; (f) Chattanooga, Tennessee, RC containing majority African American neighborhoods, some industrial areas, and the riverfront; (g) Knoxville, Tennessee, Round II EZ containing brownfield redevelopment sites, historic neighborhoods near the University of Tennessee, the riverfront convention center; and (h) Memphis, Tennessee, RC containing areas from downtown along the Mississippi River, through the Beale Street tourist district, and east to FedEx headquarters at the airport.

To preprocess the data for analysis, geographic boundary files from HUD describing the location of the RC/EZs were assembled and a 1,000-foot buffer was drawn on either side of the boundary line. These buffers were used to dummy code the businesses within 1,000 feet inside the RC/EZ and within 1,000 feet outside of the RC/EZ.³ The year that the businesses move is known at the street-level location in most places, but in some cases is only available for the census tract, ZIP code, or some other unit of geography. Observations where the locations were not known or only geocoded to the ZIP code centroid were dropped. The NETS data also included a field indicating if the organization is headquarters, branch, or subsidiary, or stand-alone establishment. As noted earlier, only establishments that are headquarters/stand-alone establishments are relevant to the issue of antipirating. The results are presented by headquarters and stand-alone establishment moves in, moves out, and net moves.

The unit of observation is the region/year where a region may be the treatment area, which is a 1,000-foot buffer inside the EZ/RC, or the control area, which is the 1,000-foot buffer outside the EZ/RC (Kolko & Neumark, 2010). See Figure 2 for an example of the buffer construction. Treatment and control regions are approximately the same size. The sample included RC/EZs listed in Figure 1. In summary, there are eight treatment areas, eight control areas, and data for 17 years⁴ (i.e., 12 years preintervention and 5 years postintervention for wage credits) for a total of 272 observations.

Figure 1.

A summary of the benefits and timeline of the EZ/RC program.

Figure 2.

The Fresno, California, EZ and location of business moves along 1,000-foot buffer.

Model Specification

The generalized linear model used to estimate adjusted interrupted time series design in this study is as follows:

\begin{array}{l} l i n k (M O V E S_{i t}) = c + d_{1} (D I M P w c_{i}) + \\ d_{2} (D I M P g r a n t_{i}) + e_{1} (D P O S T I M P w c_{i t}) + \\ e_{2} (D P O S T I M P g r a n t_{i t}) + f (T R I M P_{t}) + \\ g_{1} (T R P O S T I M P w c_{i t}) + g_{2} (T R P O S T I M P g r a n t_{i t}) + \\ h (T R A L L_{t}) + j_{1} (T R P O S T A L L w c_{t}) + \\ j_{2} (T R P O S T A L L g r a n t_{t}) + e_{i t} \end{array}

Dependent Variables

The three dependent variables in this study (MOVES_it) include the number of businesses that moved into area, the number of business that moved out of the area, and the net moves into the area. This study is able to exploit the address-level data in the NETS, which defines a move as an establishment that changes both physical address and ZIP code without returning to a prior address. In this study, openings of new establishments are not considered moves in nor are business closures considered moves out. In general, openings, closings, moves in, and moves out are a small percentage of establishment dynamics and comprise less than 10% in this sample. For most EZ/ECs, there are more openings than moves in and move closings than moves out. See Table 1 for more information. The NETS allows estimation of a change in moves consistent with the antipirating provision of the EZ/EC statute, which does not consider new establishments or new branches to be violations. The model is indexed by i, the region subscript (e.g., region within 1,000 feet of the Fresno EZ border; region outside 1,000 feet of the Fresno EZ border) and the t subscript (1990-2007).

Table 1.

Establishment Dynamics in Year 2006 as Percent of Establishments in 2007.

EZ/RC	Chattanooga, Tennessee, RC	Fresno, California, EZIII	Knoxville, Tennessee, EZII	Los Angeles, California, EZ and RC	Memphis, Tennessee, RC	San Diego, California, RC	San Francisco, California, RC	Santa Ana, California, EZII
Total number firms 2007	1,396	3,725	3,661	23,647	4,191	6,062	12,380	10,409
Net change	−12	35	−4	330	109	52	117	112
Net pct change	−0.86	0.94	−0.11	1.40	2.60	0.86	0.95	1.08
Pct openings	6.73	6.47	5.65	8.68	7.44	8.26	8.98	7.89
Pct closings	5.30	5.45	5.46	7.20	4.72	6.94	7.40	6.62
Total pct moves in	4.23	0.40	6.50	0.12	2.77	0.71	2.76	0.33
Pct moves in branch	0.43	0.05	0.79	0.01	0.10	0.08	0.09	0.06
Pct moves in HQ/standalone	3.80	0.35	5.71	0.11	2.67	0.63	2.67	0.27
Pct moves out	6.52	0.48	6.80	0.20	2.89	1.17	3.40	0.52
Pct moves out branch	0.36	0.13	0.68	0.01	0.14	0.15	0.13	0.04
Pct moves out HQ/standalone	6.16	0.35	6.12	0.19	2.74	1.02	3.27	0.48

Note. EZ = Empowerment Zone; RC = Renewal Communities; Pct = percent; HQ = headquarters. The National Establishment Time Series takes a snapshot in January each year, which is why the establishment count is from 2007, although the other data are from 2006. Only in this table, the counts include data for the entire 1990 census tract that includes a developable site, so the figures for Fresno, Santa Ana, and Knoxville are overestimated.

Treatment Variables

This model estimated the additive LATE of the RC/EZ on business relocations using neighboring areas as a control. The two levels of treatment are RC/EZs with wage credits and other tax incentives (DIMPwc_i = 1, 0 otherwise) and the Round II EZ grant (DIMPgrant_i = 1, 0 otherwise). See Figure 1 for a list of RC/EZ in each category. The treatment variables were further decomposed into the change in both the level and the trend from the entire pretreatment period to the entire posttreatment period. First, the change in level treatment effects were represented by the coefficients e₁ on dummy variable DPOSTIMPwc_it and e₂ on DPOSTIMPgrant_it that both equal 1 if the region is in an RC/EZ during the postimpact period and 0 otherwise. In other words, these are interactions of the treatment area times the treatment period. Second, the set of treatment variables representing changes in trends before and during the intervention were constructed as a vector of cardinal numbers, starting with 1 at the year during which wage credits or grants went into effect and increasing 1 per year until the last year. They were represented by coefficients g₁ on TRPOSTIMPwc_it (i.e., from 1 = 2002 to 6 = 2007, 0 otherwise) and g₂ on TRPOSTIMPgrant_it (i.e., from 1 = 2000 to 7 = 2007, 0 otherwise).

Control Variables

These variables created baseline levels and trends to support treatment effect estimation. These coefficients did not have an interesting interpretation and thus were not reported in this article, but they are available on request. For the level dummy variables, first DIMPwc_i equals 1 if they are in EZs or RCs because they all received wage credits, only at different times. Second, DIMPgrant_i served as a baseline control for the Round II EZs that received a grant (Knoxville and Santa Ana). Next, the trend variables were controlled by cardinal vectors. The variable TRALL_t indexed the trend from 1 to 17 for regions inside and outside the RC/EZ, while TRPOSTALLwc_t and TRPOSTALLgrant_t indexed the trend in all regions during the postaward period of wage credits or grants, respectively. TRIMP_t was a vector of cardinal numbers, starting at 1 for the first time period (1990 = 1) and increasing by 1 for each time period (2006 = 17). Estimates were not adjusted for selection covariates (e.g., poverty and unemployment) because I assumed that any differences that would influence business location between the treated side of the street and the other side of the street would be mean zero. One possible confounding variable would be a 1-year or multiyear recession, which in this case occurred the first year that the wage credit went into effect. As an alternate specification, a dummy variable for 2002 was included to see if this helps explain variance in the dependent variable, but it did not substantially change estimates.

Data Analysis

Longitudinal data methods known as generalized linear models with general estimating equations were used to account for correlation of repeated observations over time (Hubbard et al., 2010). The model is estimated in long data format (i.e., there are repeated observations for each time period). Furthermore, moves are count data, which means they have a lower limit of zero. Ordinary least squares regression is not an appropriate model because predicted values may fall below zero, which may violate the distributional assumptions. Count data may be modeled using a link function in the Poisson family that constrains predicted values to be zero or greater. The Poisson distribution assumes that the mean number of events per area equals the variance. Overdispersion is the condition in which the mean does not equal the variance. These data are overdispersed, so a negative binomial link function was used (Allison & Waterman, 2002; Hilbe, 2011).⁵

The coefficients for the move-in and move-out dependent variables are semielasticities, interpreted as the difference-in-differences in logs of the mean counts of the dependent variable. Because the net establishment moves are not bounded by zero, they are assumed to be continuous and modeled using panel data regression with conditional fixed effects.⁶ I first calculated pooled estimates and then estimates for each RC/EZ separately by dropping observations from other RC/EZs. It is important to note that, given the small sample sizes, the practical significance is more informative than the statistical significance. Analytic plots of the data show the trends of moves over time.⁷

Results

Descriptive Statistics

Descriptive statistics of the dependent variables for the sample may be found in Table 2. The data were decomposed by treatment assignment and treatment period. The treatment areas had higher mean establishment moves into the buffer compared with control areas within 1,000 feet of the border of the RC/EZ. However, the mean number of establishments moving out was also higher in the pretreatment area. Accordingly, prior to the implementation of the wage credits, the average net loss of establishments in the treated area was about nine establishments per RC/EZ. These moves are not normally distributed, as evidenced by the means being substantially higher than the medians. During the posttreatment period, establishment moves in and moves out increased both in the treatment and the control areas. As before, the mean number of moves in and out of the RC/EZ area was greater in the RC/EZ areas compared with adjacent areas. By the same token, the absolute value of the net loss of establishments was also greater in the treatment area. In summary, both treatment and control areas were losing establishments postintervention. The control areas, however, had relatively better relocation yields both before and after the RC/EZ intervention.

Table 2.

Descriptive Statistics of Sample.

	Pretreatment in 1,000-foot buffer						Precontrol 1,000-foot buffer outside
	Obs.	M	SD	Minimum	Maximum	Mdn	M	SD	Minimum	Maximum	Mdn
Moves in	96	47.99	56.28	3	217	17.5	38.83	43.15	0	195	21
Moves out	96	56.84	68.53	0	343	32	35.57	36.71	1	168	20
Net moves	96	−8.85	24.55	−126	34	−2	3.26	12.97	−26	48	2
Moves in solo	96	43.07	49.72	1	190	15	34.10	37.28	0	177	19
Moves out solo	96	50.30	60.14	0	296	29.5	31.71	33.02	0	150	18
Net moves solo	96	−7.23	21.68	−106	29	−2	2.40	10.82	−22	39	2
Moves in branch	96	4.88	7.44	0	32	2	4.68	6.46	0	30	2
Moves out branch	96	6.44	8.92	0	47	3	3.82	4.27	0	18	2
Net moves branch	96	−1.56	3.68	−20	5	−1	0.85	3.63	−6	17	0
	Posttreatment EZRC in 1,000-foot buffer						Postcontrol 1,000-foot buffer outside
Moves in	40	74.75	86.91	5	327	29	61.55	64.46	4	246	40
Moves out	40	93.18	102.13	7	385	51	63.68	59.76	1	215	42.5
Net moves	40	−18.43	26.71	−90	12	−6.5	−2.13	18.48	−46	53	−1.5
Moves in solo	40	69.18	80.13	4	305	27	56.35	58.32	4	229	38.5
Moves out solo	40	87.23	94.34	6	359	49	59.00	55.52	1	197	38
Net moves solo	40	−18.05	25.89	−87	11	−8.5	−2.65	17.78	−45	50	−0.5
Moves in branch	40	5.58	8.78	0	40	3	5.20	7.57	0	28	2
Moves out branch	40	5.95	9.88	0	41	2	4.68	5.16	0	19	3
Net moves branch	40	−0.38	3.35	−14	5	0	0.53	3.47	−6	9	0

Note. Obs. = Observation; EZ = Empowerment Zone; RC = Renewal Communities.

Tables 1 and 2 also decompose moves by headquarters/stand-alone (Solo) and branch establishments. As noted earlier, branch establishments comprise less than 10% of moves in this sample. Table 2 shows that net business losses increased in the RC/EZ during the wage credit period for headquarters/stand-alone businesses, but losses decreased for branches. In the control area along the border of the RC/EZ, net establishment gains into these areas decreased across both categories and remained positive. In other words, both sides of the border were worse off on a net-establishment-move basis after the wage credit intervention, but this may have been due to the 2002 recession. See Figure 3, for a plot of headquarters/stand-alone moves in (circles), moves out (triangles), and moves out (squares) by city, year, and treatment assignment. The solid vertical lines represent the smoothed net trend in relocations into the treatment area and dashed control areas. Figure 3 also has a dotted horizontal line between 1999 and 2000, when HUD awarded the Round II EZ grant, and between 2001 and 2002 to indicate the point at which the wage credits took effect. Although in the early 1990s, treatment and control areas had similar patterns, by the late 1990s, the overall trend for both moves in and moves outs had been increasing and continued to increase during the intervention period. The recession of 2002 saw move activity decrease in sample cities. As one would expect from areas targeted for economic development, these are places with a net loss in the number of establishments.

Figure 3.

Headquarters and stand-alone establishment moves in, moves out, and net moves by treatment (solid shapes) and control (hollow shapes) areas for each city.

Multivariate Statistics

Table 3 presents the results on headquarters and stand-alone establishment moves into, out of, and net moves into the RC/EZ. There was a statistically significant 0.25 increase in moves into the RC/EZ during the wage credit period beginning in 2002, holding all other variables constant.⁸ However, the slope of moves in during the wage credit period is declining and statistically significant. For the two Round II EZs, Santa Ana and Knoxville, there was a statistically significant increase in establishments moving out with an increasing trend, holding other variables constant. On the other hand, there was not a statistically significant increase in the level or trend of net moves into the RC/EZs.

Table 3.

Multivariate Results of RC/EZ Wage Credits and EZ Grants (N = 272, g = 16, t = 17).

	Moves in solo/HQ			Moves out solo/HQ			Net moves solo/HQ
Variables	Coefficient	SE	p > z	Coefficient	SE	p > z	β	SE	p > t
DIMP wage credit	0.12	0.12	.34	0.23	0.23	.33
DIMP grant	0.52	0.13	.00***	0.27	0.22	.22
Wage credit impact	0.25	0.09	.01**	0.04	0.13	.76	−9.28	7.26	.22
Grant impact	−0.13	0.08	.11	−0.23	0.10	.02***	15.33	15.86	.35
TRIMP	−0.01	0.01	.38	0.00	0.02	.90	−1.23	1.66	.47
Change in wage credit trend	−0.05	0.03	.09*	−0.02	0.04	.59	4.75	4.06	.26
Change in grant trend	0.00	0.01	.67	0.10	0.05	.03**	−2.81	2.52	.28
TRALL	0.10	0.01	.00***	0.09	0.01	.00***	0.86	0.5	.11
TRPOSTALL wage credit	−0.02	0.03	.55	−0.16	0.05	.00***	2.93	3.86	.46
TRPOSTALL grant	−0.05	0.03	.07*	0.06	0.04	.09*	−4.47	3.22	.19
Constant	1.09	0.11	.00***	1.31	0.40	.00***	−3.11	3.63	.41

Note. HQ = headquarters; RC = Renewal Communities; SE = standard error; EZ = Empowerment Zone. Moves in and moves out of stand-alone or headquarters establishments are estimated with a negative binomial population averaged model using a generalized estimating equation (xtgee in Stata) with an autoregressive correlation structure and semirobust standard errors. Fixed-effects dummies for the region (e.g., Fresno, Los Angeles, etc.) are included in those models but not displayed. Net moves are estimated using panel robust regression (xtreg command in Stata) with conditional fixed effects at the individual level.

p < .1. **p < .05. ***p < .01 (statistically significant codes).

Finally, Table 4 shows results for specific RC/EZs. This allows comparison of EZ with RC to provide insight into the prohibition against business relocation. All RC/EZs saw net increases except Los Angeles,⁹ San Diego, and San Francisco. The San Diego RC saw the only statistically significant increase in net establishment moves, with a net loss of 56.81 establishments, compared with the adjacent area and the pretreatment period. However, the postintervention trend in San Diego was a net positive of 12.45 establishments per year per area. Based on an inspection of Figure 3, it appears that San Diego may be making up losses from the 2002 recession compounded by its tourism-heavy economy. In a similar vein, San Francisco’s trend increased in net moves by 36.41 establishments per area per year, compared with the control area and pretreatment period. Table 4 also shows individual city-level regressions for headquarters and stand-alone moves in and moves out. The Fresno EZ, Los Angeles RC/EZ, and Chattanooga RC each had statistically significant increases in businesses moving into the zone. The effect sizes for changes in trends for moves in are small and not statistically significant.¹⁰

Table 4.

Local Average Treatment Effects by City on Stand-Alone or Headquarters Firms (Baseline Dummies and Trend Variables not Reported; n = 34, g = 2, t = 17).

		Moves in solo			Moves out solo			Net moves solo
City	Label	Estimate	SE	p > \|t\|	Estimate	SE	p > \|t\|	Estimate	SE	p > \|t\|
Fresno EZ	Period	0.78	0.32	.02**	0.37	0.29	.20	3.96	5.96	.51
Fresno EZ	Trend	−0.10	0.11	.36	−0.09	0.10	.37	0.51	2.05	.81
Los Angeles RC/EZ	Period	0.38	0.20	.07*	0.49	0.18	.01**	−14.59	20.97	.49
Los Angeles RC/EZ	Trend	−0.08	0.07	.26	−0.02	0.06	.69	−5.83	7.21	.43
San Diego RC	Period	0.26	0.20	.20	0.81	0.17	.00***	−56.81	15.28	.00***
San Diego RC	Trend	−0.17	0.07	.02**	−0.24	0.07	.00***	12.45	5.25	.03**
San Francisco RC	Period	0.10	0.19	.58	0.16	0.19	.41	−18.36	30.19	.55
San Francisco RC	Trend	0.05	0.06	.41	−0.14	0.07	.03**	36.41	10.38	.00***
Santa Ana EZ wage	Period	0.18	0.26	.50	0.21	0.26	.41	−14.80	20.39	.48
Santa Ana EZ wage	Trend	−0.05	0.31	.88	−0.55	0.43	.20	8.22	22.27	.72
Santa Ana EZ grant	Period	−0.37	0.47	.44	−0.08	0.30	.79	19.38	32.52	.56
Santa Ana EZ grant	Trend	−0.08	0.31	.79	0.01	0.28	.97	−10.57	21.31	.63
Chattanooga RC	Period	0.79	0.44	.08*	0.57	0.52	.27	1.77	6.24	.78
Chattanooga RC	Trend	0.03	0.16	.86	0.00	0.18	1.00	−1.88	2.15	.39
Knoxville EZ wage	Period	0.15	0.41	.71	−0.38	0.35	.28	3.20	4.95	.52
Knoxville EZ wage	Trend	0.88	0.43	.04**	−0.10	0.48	.84	10.64	5.41	.06*
Knoxville EZ grant	Period	0.83	0.65	.20	0.32	0.75	.67	7.40	7.90	.36
Knoxville EZ grant	Trend	−0.87	0.44	.05*	0.08	0.47	.87	−11.04	5.18	.04**
Memphis RC	Period	0.35	0.34	.30	−0.28	0.42	.51	7.79	9.32	.41
Memphis RC	Trend	−0.02	0.11	.83	0.13	0.13	.31	−2.24	3.20	.49

Note. SE = standard error; EZ = Empowerment Zone; RC = Renewal Communities. Moves in and moves out are estimated with a negative binomial population averaged model using a generalized estimating equation with an autoregressive correlation structure. Dummies for the treatment region and trend baseline variables are included in those models but not displayed. Net moves are estimated using panel regression (xtreg command in Stata) with conditional fixed effects at the individual level.

p* < .1. p** < .05. p*** < .01 (statistically significant codes).

A review of EZ annual reports on HUD’s website provides insight into the results (U.S. Department of Housing and Urban Development, 2013). The downward establishment trend in Knoxville may be the result of including the EZ long-term public works projects funded by other sources (e.g., major brownfield redevelopment, convention center). These projects may not have had private sector tenants show up in the analysis. In contrast, the Santa Ana EZ emphasized workforce development and business expansion within existing sites. Fresno’s significant increase in moves into the EZ is consistent with its aggressive brownfield redevelopment strategy of areas near Yosemite airport.

Discussion and Conclusion

The motivation behind this study was to test a concern in the policy literature that a geographically targeted benefit, in particular a tax incentive, would result in the relocation of establishments. Given the strength of the design, it is reasonable to conclude that observed relocation changes over and above previous trends and comparison areas were associated with the program incentives. Although there is evidence that the wage credit stimulates business moves into at least some of the RC/EZ areas in this sample (e.g., Fresno, Los Angeles, Chattanooga), the wage credit also appears to stimulate business moves out some of the RC/EZ (Los Angeles, San Diego), resulting in no statistically significant net changes in moves except in San Diego. These results may be explained by Hanson and Rohlin (2011b), who find that incoming establishments that are able to take advantage of wage credits, such as those that are labor intensive, outbid those that are not leading to an increase in service and retail establishments, while transportation, financial, insurance, and real estate sectors decrease.

With regard to the prohibition against relocation, one EZ, one RC, and one city all saw increases in business relocation. However, the two cities with EZ grants did not see a significant increase in business relocations. This could be evidence that the prohibition worked; however, these data also show that these RC/EZ programs may drive out more businesses out than they attract. The fact that the prohibition against business relocation applied only to the direct grants of the EZ program and not to the tax incentives may require further thought if Congress plans to authorize more spatially targeted tax incentive programs. It is not necessarily warranted to discuss an antidisplacement policy because policy makers may be more concerned with the changes in employment than number of establishments and be willing to accept a sorting that yields more jobs. Although it may be ethical to prohibit public assistance for business relocation, it may be naive to think that retail, commercial, or industrial vacancies can be filled only by branches and affiliates of existing establishments, because they comprise less than 10% of establishments. If Congress is still concerned about antipirating, it may wish to extend tax relief only to incumbent establishments for expansion of the workforce or to new businesses. Congress may also wish to provide incentives for new hires at existing establishments.

Limitations

The major limitation of this study stems from the fact that the federal government did not assign the place-based initiatives randomly; therefore, an average treatment effect could not be properly estimated. The second limitation is that the ideal counterfactual to distinguish the effect of the antipirating policy from the relocation effect of the EZ program inputs would be to have an EZ program that was the same in all aspects except the antipirating provision. This study could only compare the EZ and RC programs to control areas and the RC program, and although the program had similar tax incentives, it did not have grants and therefore was not an ideal counterfactual. The third major limitation is that previous literature on EZs documents a potential negative spillover effect from using a counterfactual adjacent to the treatment area. Because the estimate of the treatment effects on net relocation are not statistically significant, I am comfortable that the estimate, if biased upward, did not produce a false positive. However, the net move estimates could be a false negative if the true spillovers for relocation are positive in these samples and drove relocations to the control areas. Indeed, six of the eight RC/EZs had control areas with more net moves in than the treatment areas in the treatment period. Further research would have to distinguish spillovers using comparison areas that are not adjacent.

There are also data limitations that stem from using self-reported business data because the nonresponse rate is unknown and may not be missing at random. It is unlikely, however, that the nonresponse rate would vary substantially across the street. This study also suffered from incomplete geocoding records. About 30% of the records in NETS had been assigned latitudes and longitudes that were the centers of ZIP codes rather than the actual locations, and those that could not be regeocoded were dropped. Also, pooling the treatment effect of EZs and RCs implicitly averages differences between the smaller monetary value of incentives in the programs. Another limitation is that the sample size is small because of budget constraints. Finally, the treatment and control areas are approximately the same size because they are 1,000 ft from either side of the center of a street. The interior areas, however, are systematically smaller because they pick up the opposite corner of an intersection. For the exterior corners, the buffer contains three corners of an intersection. This would bias the calculation downward near larger control areas; however, some designated areas have a “Swiss cheese” configuration that contains interior corners.

These limitations notwithstanding, this study adds to the existing literature and provides insights regarding the relocation of businesses into areas with geographically targeted tax incentives. Because the RC/EZ program has ended, further research is warranted to determine whether or not the expiration of the incentives harmed businesses or caused them to shed employees. To determine the quality of jobs created or lost and the impact on the tax base, further research could also to be done to identify the kinds of businesses that move in and move out of the RC/EZs. Given the complexity of urban areas, future programs need to be carefully structured.

Footnotes

Acknowledgements

I would like to thank panelists and audience members at the Association for Public Policy Analysis and Management for their helpful comments.

Authors’ Note

A version of this article was presented at the 33rd annual Association for Public Policy Analysis and Management (APPAM) research conference held in Washington, D.C., November 3 to 5, 2011. The article benefited from comments from Al Acker, George Galster, Danielle LeVaque-Manty, Philip Tegeler, and Chris Walker. Conclusions drawn in this article are solely those of the author and not of the Department of Housing and Urban Development.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Support for this study came from the Fisher Center for Real Estate and Urban Economics and the U.S. Department of Housing and Urban Development Doctoral Dissertation Research Grant (H-21569 SG).

Notes

Author Biography

Richard Smith is an assistant professor of social work at Wayne State University. His research interests include sustainable community development, inequality, and migration. Smith has published in Social Work, Letters in Spatial and Resource Sciences, and Urban Studies.

References

ADP. (2010). Federal and state tax credits and incentives. Retrieved from http://www.adp.com/solutions/employer-services/tax-and-compliance/large-business/tax-credits-and-incentives/federal-and-state-tax-incentives.aspx

Allison

P. D.

Waterman

R. P.

(2002). Fixed–effects negative binomial regression models. Sociological Methodology, 32, 247-265. doi:10.1111/1467-9531.00117

Arauzo-Carod

J.-M.

Liviano-Solis

Manjón-Antolín

(2010). Empirical studies in industrial location: An assessment of their methods and results. Journal of Regional Science, 50, 685-711. doi:10.1111/j.1467-9787.2009.00625.x

Bartik

(1991). Who benefits from state and local economic development policies? Retrieved from http://research.upjohn.org/up_press/77

Brouwer

A. E.

Mariotti

van Ommeren

J. N.

(2004). The firm relocation decision: An empirical investigation. Annals of Regional Science, 38, 335-347. doi:10.1007/s00168-004-0198-5

Busso

Gregory

Kline

P. M.

(2010). Assessing the incidence and efficiency of a prominent place based policy (Working paper No. 16096). Retrieved from http://www.nber.org/papers/w16096

Busso

Kline

(2008). Do local economic development programs work? Evidence from the federal empowerment zone program. Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1090838

Chaskin

R. J.

Peters

C. M.

(1997). Governance in empowerment zone communities: A preliminary examination of governance in fifteen empowerment zone communities. Retrieved from https://www.chapinhall.org/sites/default/files/CB_10.pdf

First Advantage. (2010). Location-based incentives programs. Retrieved from http://www.fadv.com/Location-Based-Incentives.aspx

10.

Galster

Temkin

Walker

Sawyer

(2004). Measuring the impacts of community development initiatives: A new application of the adjusted interrupted time-series method. Evaluation Review, 28, 502-538. doi:10.1177/0193841X04267090

11.

Gittell

Newman

(1998). Empowerment Zone implementation: Community participation and community capacity. Retrieved from http://www.ebmc.org/home/documents/Packet%209.pdf

12.

Gittell

Newman

Bockmeyer

Lindsay

(1998). Expanding civic opportunity: Urban empowerment zones. Urban Affairs Review, 33, 530-558. doi:10.1177/107808749803300408

13.

Gittell

Newman

Pierre-Louis

(2001). Empowerment Zones: An opportunity missed: A six-city comparative study. Retrieved from https://www.gc.cuny.edu/CUNY_GC/media/CUNY-Graduate-Center/PDF/Centers/Howard%20Samuels/Empowerment_Zones_Study.pdf

14.

Halpern

(1995). Rebuilding the inner city. New York, NY: Columbia University Press.

15.

Ham

J. C.

Swenson

Imrohoroglu

Song

(2011). Government programs can improve local labor markets: Evidence from state enterprise zones, federal empowerment zones and federal enterprise community. Journal of Public Economics, 95, 779-797. doi:10.1016/j.jpubeco.2010.11.027

16.

Hanson

(2009). Local employment, poverty, and property value effects of geographically-targeted tax incentives: An instrumental variables approach. Regional Science and Urban Economics, 39, 721-731. doi:10.1016/j.regsciurbeco.2009.07.002

17.

Hanson

(2011). Utilization of employment tax credits: An analysis of the empowerment zone wage tax credit. Public Budgeting & Finance, 31(1), 23-36. doi:10.1111/j.1540-5850.2011.00972.x

18.

Hanson

Rohlin

(2011a). Do location-based tax incentives attract new business establishments? Journal of Regional Science, 51, 427-449. doi:10.1111/j.1467-9787.2010.00704.x

19.

Hanson

Rohlin

(2011b). The effect of location-based tax incentives on establishment location and employment across industry sectors. Public Finance Review, 39, 195-225. doi:10.1177/1091142110389602

20.

Hanson

Rohlin

(2013). Do spatially targeted redevelopment programs spillover? Regional Science and Urban Economics, 43, 86-100. doi:10.1016/j.regsciurbeco.2012.05.002

21.

Hebert

Vidal

Mills

James

Gruenstein

(2001). Interim assessment of the empowerment zones and enterprise communities (EZ/EC) program: A progress report. Washington, DC: U.S. Department of Housing and Urban Development. Retrieved from www.huduser.org/Publications/pdf/ezec_Report.pdf

22.

Hetzel

O. J.

(1994). Some historical lessons for implementing the Clinton administration’s empowerment zones and enterprise communities program: Experiences from the model cities program. Urban Lawyer, 26, 63-81.

23.

Hilbe

J. M.

(2011). Negative binomial regression. New York, NY: Cambridge University Press.

24.

Hubbard

A. E.

Ahern

Fleischer

N. L.

Laan

M. V.

Lippman

S. A.

Jewell

. . . Satariano

W. A.

(2010). To GEE or not to GEE: Comparing population average and mixed models for estimating the associations between neighborhood risk factors and health. Epidemiology, 21, 467-474. doi:10.1097/EDE.0b013e3181caeb90

25.

Kolko

Neumark

(2010). Do some enterprise zones create jobs? Journal of Policy Analysis and Management, 29, 5-38. doi:10.1002/pam.20477

26.

Krupka

D. J.

Noonan

D. S.

(2009a). Empowerment Zones, neighborhood change and owner-occupied housing. Regional Science and Urban Economics, 39, 386-396. doi:10.1016/j.regsciurbeco.2009.03.001

27.

Krupka

D. J.

Noonan

D. S.

(2009b). Neighborhood dynamics and the housing price effects of spatially targeted economic development policy (Discussion Paper No. 4308). Retrieved from ftp://ftp.iza.org/SSRN/pdf/dp4308.pdf

28.

Lavin

Whysall

(2004). From enterprise to empowerment: The evolution of an Anglo-American approach to strategic urban economic regeneration. Strategic Change, 13(4), 219-229. doi:10.1002/jsc.681

29.

Lee

(2005). The Federal Empowerment Zone and Enterprise Community Program: Outcome on labor income. Retrieved from https://papyrus.bib.umontreal.ca/xmlui/handle/1866/283

30.

Levine

D. I.

(1999). Gilding the ghetto. Across the Board, 36(9), 10.

31.

Marcuse

(1997). The enclave, the citadel, and the ghetto: What has changed in the post-Fordist U.S. city. Urban Affairs Review, 33(2), 228-264. doi:10.1177/107808749703300206

32.

Morgan

S. L.

Winship

(2007). Counterfactuals and causal inference: Methods and principles for social research (1st ed.). New York, NY: Cambridge University Press.

33.

Nathan

R. P.

Wright

D. J.

(1997). Empowerment Zone initiative: Building a community plan for strategic change findings from the first round of assessment. Albany, NY: State University of New York, The Nelson A. Rockefeller Institute of Government.

34.

Neumark

Zhang

Wall

(2005). Employment dynamics and business relocation: New evidence from the National Establishment Time Series (NBER Working Paper No. W11647). Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=819807

35.

Oakley

Tsao

H.-S.

(2006). A new way of revitalizing distressed urban communities? Assessing the impact of the Federal Empowerment Zone program. Journal of Urban Affairs, 28, 443-471. doi:10.1111/j.1467-9906.2006.00309.x

36.

O’Connor

(1999). Swimming against the tide: A brief history of federal policy in poor communities. In Ferguson

R. F.

Dickens

W. T.

(Eds.), Urban problems and community development (pp. 77-137). Washington, DC: Brookings Institution Press.

37.

Office of Management and Budget. (2011). The President’s budget for fiscal year 2012. Retrieved from https://www.whitehouse.gov/files/documents/budget_2012.pdf

38.

Office of the Press Secretary. (2013). Fact sheet: The President’s plan to ensure hard work leads to a decent living. Retrieved from http://www.whitehouse.gov/the-press-office/2013/02/15/fact-sheet-president-s-plan-ensure-hard-work-leads-decent-living

39.

Omnibus Budget Reconciliation Act of 1993, H.R.2264 (1993). Retrieved from http://www.gpo.gov/fdsys/pkg/BILLS-103hr2264enr/pdf/BILLS-103hr2264enr.pdf

40.

Quigley

J. M.

(1994). New directions for urban policy. Housing Policy Debate, 5, 97-106. doi:10.1080/10511482.1994.9521153

41.

Quigley

J. M.

(2000). A decent home: Housing policy in perspective. Brookings-Wharton Papers on Urban Affairs, 2000, 53-88. doi:10.1353/urb.2000.0011

42.

Rich

M. J.

(1989). Distributive politics and the allocation of federal grants. American Political Science Review, 83, 193-213. doi:10.2307/1956440

43.

Rich

M. J.

Stoker

R. P.

(2010). Rethinking empowerment: Evidence from local empowerment zone programs. Urban Affairs Review, 45, 775-796. doi:10.1177/1078087410366530

44.

Robinson

J. A.

(2005). An economic sociology of entry barriers: Business entry and the inner city market. New York, NY: Columbia University.

45.

Rosenbaum

P. R.

(2002). Observational studies (2nd ed.). New York, NY: Springer.

46.

Rubin

H. J.

(1988). Shoot anything that flies; claim anything that falls: Conversations with economic development practitioners. Economic Development Quarterly, 2, 236-251. doi:10.1177/089124248800200304

47.

Schultz

(2011, October 22). Altered “9-9-9” proposal has exemptions for poor. The Detroit News. Retrieved from http://www.news-databasex.blogspot.in/2011/10/altered-9-9-9-proposal-has-exemptions.html

48.

Smith

R. J.

(2007). Real estate capitalization by policy input in the Cleveland empowerment zone: An application of adjusted interrupted time series analysis (AITS). Michigan Journal of Public Affairs, 4, 1-22. Retrieved from http://mjpa.umich.edu/files/2014/06/2007-Smith-ClevelandRealEstate.pdf

49.

Smith

R. J.

(2011). Immigration, spatial mismatch and entrepreneurship in empowerment zones and enterprise communities. Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1837668

50.

Smith

R. J.

(2015). Empowerment for whom? The impact of community renewal tax incentives on jobs and businesses. Urban Studies, 52, 702-720. doi:10.1177/0042098014528398

51.

Taxpayer Relief Act of 1997, 26 U.S.C. § 1391 to 1397 (1997). Retrieved from http://www.gpo.gov/fdsys/pkg/USCODE-2009-title26/html/USCODE-2009-title26-subtitleA-chap1-subchapU.htm

52.

U.S. Department of Housing and Urban Development. (1995). Urban empowerment zones and enterprise communities: Round one designations, 24 C.F.R. § 597. Retrieved from http://www.gpo.gov/fdsys/pkg/CFR-2013-title24-vol3/pdf/CFR-2013-title24-vol3-part597.pdf

53.

U.S. Department of Housing and Urban Development. (1998). Urban empowerment zones: Round two and three designations, 24 C.F.R. § 598. Retrieved from http://www.gpo.gov/fdsys/pkg/CFR-2012-title24-vol3/pdf/CFR-2012-title24-vol3-part598.pdf

54.

U.S. Department of Housing and Urban Development. (2001). Renewal communities, 24 C.F.R. § 599. Retrieved from http://www.gpo.gov/fdsys/pkg/CFR-2013-title24-vol3/pdf/CFR-2013-title24-vol3-part599.pdf

55.

U.S. Department of Housing and Urban Development. (2010). EZ/RC employment credit claims for processing years 1997-2008. Retrieved from http://portal.hud.gov/hudportal/documents/huddoc?id=19170_employmentcreditM.pdf

56.

U.S. Department of Housing and Urban Development. (2011). RC/EZ initiative timeline. Retrieved from http://www.hud.gov/offices/cpd/economicdevelopment/programs/rc/about/timeline.cfm

57.

U.S. Department of Housing and Urban Development. (2013). Welcome to the community renewal initiative. Retrieved from http://portal.hud.gov/hudportal/HUD?src=/program_offices/comm_planning/economicdevelopment/programs/rc

58.

U.S. Government Accountability Office. (1997). Economic development activities: Overview of eight federal programs (No. RCED-97-193). Retrieved from http://www.gao.gov/assets/230/224739.pdf

59.

U.S. Government Accountability Office. (1999). Community development: Businesses’ use of empowerment zone tax incentives (No. RCED-99-253). Retrieved from http://www.gao.gov/assets/230/228056.pdf

60.

U.S. Government Accountability Office. (2004). Community development: Federal revitalization programs are being implemented, but data on the use of tax benefits are limited (No. GAO-04-306). Retrieved from http://www.gao.gov/assets/250/241656.pdf

61.

U.S. Government Accountability Office. (2006a). Empowerment zone and enterprise community program: Improvements occurred in communities, but the effect of the program is unclear (No. GAO-06-727). Retrieved from http://www.gao.gov/assets/260/251626.pdf

62.

U.S. Government Accountability Office. (2006b). Empowerment zone and enterprise community program: Survey of round I federal enterprise communities (No. GAO-06-734SP). Retrieved from http://www.gao.gov/assets/210/202969.pdf

63.

U.S. Government Accountability Office. (2007, September). Economic development: Formal monitoring approaches needed to help ensure compliance with restrictions on funding employer relocations (No. GAO-07-1005). Retrieved from http://www.gao.gov/assets/270/266586.pdf

64.

U.S. Government Accountability Office. (2010). Revitalization programs: Empowerment zones, enterprise communities, and renewal communities (No. GAO-10-464R). Retrieved from http://www.gao.gov/products/GAO-10-464R

65.

Wallace

M. A.

(2003). An analysis of presidential preferences in the distribution of empowerment zones and enterprise communities. Public Administration Review, 63, 562-572. doi:10.1111/1540-6210.00320

66.

Wallace

M. A.

(2004). Congressional considerations and urban characteristics in the selection of empowerment zones and enterprise communities. Journal of Urban Affairs, 26, 593-609. doi:10.1111/j.0735-2166.2004.00216.x

67.

Walls

D. W.

(2008). National Establishment Time-Series (NETS) Database: Data description. Retrieved from https://msbfile03.usc.edu/digitalmeasures/cswenson/intellcont/NETS%20Database%20Description2008-1.pdf