Increased access to capital: evaluation of the New Market Tax Credit Program in New York

PDF

Download

Open document

Flip pages

Summarize

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content 1 Increased Access to Capital Evaluation of the New Market Tax Credit Program in New York By: Jacqueline Berman BS Finance and Economics University of Illinois at Chicago, 2006 Submitted to the Department of Economics in partial fulfillment of the requirements for a Degree of Master of Economics and Urban Planning At The University of Southern California May 2014 2 ABSTRACT The Community Renewal Tax Relief Act of 2000 was enacted to spur development in economically distressed communities within the United States. One of provisions of this bill was to create tax incentives for investment in small businesses within these distressed areas. This provision led to the creation of the federal economic development program, New Market Tax Credit Program, which draws private investors to low income neighborhoods through the provision of federal tax credits. Originally, the Program was approved for 7 years and has been extended in subsequent years. In 2013, this program came into question of whether or not it should be extended once again and whether it was working. However, supporting documentation of the impact of the program is sparse. Since the implementation of this program in 2003, there have been gaps in its program assessment, which this paper attempts to fill. This paper adopts both case study and quantitative methods to analyze the impact of the New Market Tax Credit Program on census tracts within New York. A difference in difference regression model was used to compare changes in poverty rates between low income census tracts that have received investment due to the Program and those have not and estimate the effect of the Program. Using the case study method, projects funded using New Market Tax Credits showed small employment growth and greater access to health and human services in high need neighborhoods. The results of the econometric analysis show that there is a statistically significant negative correlation between receiving investment due to the Program and the change in poverty rate. THESIS ADVISING COMMITTEE: Geert Ridder, Professor in the Department of Economics Joel David, Associate Professor in the Department of Economics Leonard Mitchell, Clinical Professor in the Department of Planning 3 TABLE OF CONTENTS LIST OF ACRONYMS 4 LIST OF FIGURES 5 LIST OF TABLES 6 Introduction 7 Thesis Organization The Significance of the New Market Tax Credit Program Chapter One 13 The Need for Intervention Overview of Local Economic Development History of Local Based Initiatives Overview of New Market Tax Credit Program Previous Literature Chapter Two 30 Three Case Studies Case Study Comparison Chapter Three 43 Data Methodology Results Chapter Four 59 Conclusion Policy Recommendations Bibliography 64 Appendix 68 4 ACRONYMS ACS American Community Survey ATE Average Treatment Effect ATET Average Treatment Effect on the Treated Individuals CDE Community Development Entity CDFI Community Development Financial Institution CIIS Community Investment Impact System CNMC Chase New Markets Corporation CRA Community Reinvestment Act CUCS Center for Urban Community Services DID Difference In Difference ERF Empowerment Reinvestment Fund GAO Government Accountability Office HUD U.S. Department of Housing and Urban Development LISC Local Initiatives Support Corporation NFF Nonprofit Finance Fund NMSC New Markets Support Company NMTC New Market Tax Credit NDC National Development Council OECD Organization for Economic Cooperation and Development OLS Ordinary Least Squares QEI Qualified Equity Investments QLICI Qualified Low-Income Community Investment QALICB Qualified Active Low-Income Community Business ROI Return on Investment SBA U.S. Small Business Administration UI Urban Institute 5 LIST OF FIGURES Figure 1 New Market Tax Credit Capital Flow Chart Figure 2 Example of a Leveraged New Market Tax Credit Transaction Figure 3 Program Evaluation Framework Figure 4 New Market Tax Credit Allocations in New York Figure 5 Distribution of Poverty Rates within New York 6 LIST OF TABLES Table 1 Characteristics of Selected CDEs Table 2 Characteristics of Selected Projects Table 3 Names, Description, and Sources for Variables Table 4 Comparison of the Poverty Rate between Groups Table 5 Characteristics of New York Census Tracts Table 6 Reduced Form Models for Change in Poverty Rate Table 7 Reduced Form Models for Change in Employment Table 8 Reduced Form Models for Change in Migration Table 9 Comparison of Standard Errors 7 Introduction “The jury is still out on whether economic development policies have any effect at all.” - Therese J. McGuire (1992) The Unites States Congress passed the New Market Tax Credit Program (NMTC) in 2000 to encourage new and additional investment into low-income communities. For the purpose of this program, low-income communities are defined as a census tract with a poverty rate of 20 percent or higher, a nonmetropolitan census tract with a median family income that does not exceed 80 percent of the statewide family income, a census tract within a metropolitan area with family median income that is no greater than 80 percent of the statewide or metropolitan family income, a census tract with low population, a rural census tract with high migration, or certain areas that are not included within a census tract (CDFI, 2013). The NMTC program was part of the Clinton administration’s Community Renewal Tax Relief law, which was aimed to reduce poverty through economic growth (New Markets Tax Credit Coalition, 2007). The idea behind the NMTC legislation is that there are viable business opportunities within low income communities, but the cost and availability of capital in these ‘New Markets’ stunt economic growth. This economic development initiative was to lure both individuals and corporations with tax credits in exchange for investing in operating businesses and real estate projects located in low-income neighborhoods. The NMTCs are distributed by the U.S. Department of the Treasury’s Community Development Financial Institutions (CDFI) Fund to community development entities (CDEs) for projects within low income communities. CDEs then entice investors for these projects with tax credits equivalent to 39% of the initial investment. Tax credits are dispersed over a 7 year period, where investors are given 5% of their initial investment for the first three years and 6% for the last four years. “The purpose of the legislation was to stimulate private sector investment in urban and rural low income communities and build a delivery system of private for-profit and non-profit entities that could provide technical and financial assistance to economically distressed urban and rural communities and their businesses,”(New Markets Tax Credit Coalition, 2007, pp. 2). While national economic development policies focusing on specific locations date back to the New Deal, much has yet to be done to effectively address communities left behind in America’s 8 economic expansion. The NMTC program may be the economic policy that is needed to start targeting these communities. The program sets itself apart from previous attempts at spurring economic development by blending local based incentives with private market investment models. Further this program utilizes private investment models to direct capital to low-income communities. In the past, the federal government’s solution to invigorating economic development was to provide physical infrastructure to spur development opposed to directly intervening in with the private sector (Blakely & Leigh, 2010). Due to lack of effective national policy to address job creation and the limitation of national economic policy to tackle both the regional and local employment requirements and economic sectoral adjustments, the federal government and local officials acknowledge the best place for the national government should be more supportive financially rather than asserting their power to force economic change. The issue now is that federal funding is dwindling while the need for federal assistance is growing at the same time. In response, the federal government has tried to make it easier to access the remaining funds by cutting the red tape. This has allowed local officials to gain greater control of the resources available; however, the same problem persists. More funding is needed to solve current economic issues than is available. The NMTC program specifically addresses this issue by leveraging the federal funds left with private investment. Rather than reinvent the wheel, the federal government essentially relies on the CDEs for their expertise in raising capital to follow the existing venture capital model. By recognizing the lack of access to funding to low-income communities and directing policy at this issue, the program has hope to direct large amounts of capital given the limited funding using methods that have a history of working in the private sector. The NMTC program’s location specific component makes this economic development tool special. The program narrows the scope of economic development to target low-income communities that do not have the same accessibility to resources that wealthier communities have. By narrowing the focus to target low income communities, this policy can be categorized as community economic development which is defined as “(1) efforts to develop housing, jobs or business opportunities for low-income people, (2) in which a leading role is played by non-profit, nongovernmental organizations, (3) are accountable to residentially defined communities.” Another key aspect of community economic development is local ownership and control.. Local ownership boosts the economy in multiple ways. One way is that the income earned through local owners is more likely to stay and be spent within the community. Another is that local 9 owners are likely to know internal needs as they are a part of the community. Lack of capital within the community limits the number of local owners. The NMTC program addresses the issue of ownership by not interfering with the projects themselves but rather directs investments to neighborhoods overlooked by traditional financial institutions. As the end of 2012, $31 billion in credit had been dispensed by the CDFI. As of April 2013, Congress has approved $40 billion in NMTCs. This includes: $15 billion from the Community Renewal and Tax Relief Act, $1 billion from the Gulf Opportunity Zone Act, $3.5 billion from the Tax Relief and Health Care Act of 2006, an additional $3.5 from the Tax Relief and Health Care Act of 2006, $3.5 billion from the Emergency Economic Stabilization Act in 2008, $5 billion from American Recovery and Reinvestment Act, $3.5 billion from the 2010 Tax Relief, Unemployment Insurance Reauthorization and Job Creation Act, and $3.5 billion from the American Taxpayer Relief Act . While the nominal rate of return is 39% of the investment, the effective rate is 26% including the loss of revenue and government’s cost. Even though this policy is well intentioned and addresses issues that were not previous considered, has the government spent this money wisely? We must take a critical look at whether or not this program has made a considerable effect on communities allocated NMTCs. The question being explored is: Has the New Market Tax Credit Program made a significant impact on poverty in low-income communities in New York? This paper will examine impact of the NMTC program on the percentage of those in poverty within New York census tracts that have been allocated NMTCs. This thesis argues two main points. First, economic development policies should be evaluated with impacts measured to ensure federal funds are used efficiently. Many articles available praise the program as being effective in raising capital for low-income areas, but this thesis argues whether or not we are seeing 10 significant impacts that warrant the extension of the program or whether we need to adjust this method to see a greater change. Secondly, this thesis presents the idea that there is minimal social impact resulting from the NMTC program, which is shown through both case studies and an econometric analysis. Thesis Organization Chapter One will discuss the process of the New Market Tax Credit Program and provide context to why the program was constructed the way it was. First, this chapter will provide a brief description of the demand for local economic development and the need for intervention. Next, economic development policy including economic theory supporting government intervention will be discussed. Historical trends of economic development policy will be included to show the progression of these types of policies. This chapter will then detail the NMTC program and its process. Chapter Two will provide a brief case study analysis of five projects within New York that received investment due to NMTCs. This chapter will provide insight to the ways that previous government entities and organizations have analyzed the program as well as provide a comparison of results from previous reports and the analysis discussed in this paper. The projects studied were chosen based on the success of the associated Community Development Entity and information availability. The projects are then compared to each other to find patterns across successful projects, which is the method most consistent with previous literature on the Program. Chapter Three will introduce the method of analysis. First, this chapter will begin with providing the reasoning and theory on the statistical model used. This will include the rationale for how both independent and dependent variables were chosen for analysis, based on political views of social improvement as well as econometric constraints. Second, this chapter will then detail the sources of data collection and descriptive statistics. Third, this chapter will present an analysis of the results of the regression detailed previously. Fourth, this chapter will discuss the advantages and disadvantages of the model used in the regression. Fifth, this chapter will provide 11 a new perspective for analyzing economic development policy including my argument that analyzing the economic development policies and measuring social impacts will provide a better platform to distinguish between effective and ineffective policies. Finally, this chapter will also compare the case study method used by many to a statistical approach to show the need for a more rigorous examination of federally funded economic development programs. Chapter Four will provide a conclusion and policy recommendations. This chapter will also discuss the need for further analysis of the NMTC program in the future to evaluate the social impacts of the program in the long term. I will address the two key issues the thesis has addressed; economic development policies should be evaluated with impacts measured to ensure federal funds are used efficiently and that there is minimal social impact resulting from the NMTC program. This chapter concludes with a call to more rigorously measure the impacts of economic development policies and provide policy recommendation based on outcomes for continued economic development growth in low-income communities. The Significance of Studying the New Market Tax Credit Program Last year, the July 26, 2013 tax reform proposal deadline was reported in newspapers across the country. With the US government looking to reevaluate existing tax credit programs, determining which ones are effective are crucial to the future of our economy. Examining this issue will provide greater insight to what the future of this Program should be surrounding issues measuring its impact. By identifying the Program’s impact, we can start the conversation of whether or not continued financing for this Program is worth it. Economic development regardless of location requires capital. Unfortunately, low income communities tend to find it more difficult to attract the required capital. This analysis of the NMTC Program furthers the discussion of how to evaluate economic development programs and to what extent these programs address poverty in low income neighborhoods. While the NMTC Program is a hybrid of different strategies, it leverages federal funds with private capital with a place based approach. The program possesses a unique combination of strategies in its approach. It combines local base incentives with the New Markets Model, which focuses on funneling capital into specific areas. By examining the impact of the program, we can adjust government funded economic development to be more effective in alleviating poverty. 12 Secondly, the importance of this paper is to highlight the need for evaluating economic development policies. The bigger picture of analyzing the NMTC program is to determine whether it can be used to effectively increase economic activity in areas with a high concentration of poverty by using multiple methods of analysis. Continued evaluations of economic policies will contribute to the discussion of how best to assess economic development programs so that we can utilize public funding in an effective matter rather than out of goodwill alone. The majority of existing literature specific to the NMTC Program includes case studies and comparing observations of the NMTC funded projects rather than econometric models quantifying its overall impact. The Government Accountability Office has examined the implementation of the program, but little work has been done to analyze the specific impact of the NMTC Program on neighborhood characteristics, such as the poverty or unemployment rate. While economic initiatives in the past have attempted to alleviate poverty, the NMTC Program provides a new take. The NMTC has allowed the federal government to shine the light on community development venture capital, which allows for a “double bottom line,” where investors expect both a financial and social returns. This program seeks to connect investors to job and wealth creation within low-income communities through private sector models. The NMTC Program directs investors to projects within areas that may have been passed up by offering financial incentives for the extra risk or perceived risk rather than having them invest in already well off neighborhoods. This program may prove to be a successful incentive to address the problem of how to create wealth in low income areas. By taking a closer look at the impacts of the program on qualified neighborhoods, we can move closer to alleviating poverty within the country and possibly extend the application to other developed countries. The reason we are looking at this program in the context of New York is that New York has a high proportion of low income communities and encompasses the largest urban population within the United States (Central Intelligence Agency, 2013). The most recent U.S. Census Bureau’s estimates the total population of New York in 2012 to be 19,570,261 individuals with 8,336,697 or 42.598% living in New York City. One of the defining characteristics of New York City is high density. Using the U.S. Census Bureau’s 2010 data, New York City had a population density of 27,012.5 people per square miles compared to New York State’s at 411.2 and the entire country’s population density of 87.4 people per square mile. This attribute allows for New 13 York to be used as a model for other urban areas, both current and future, as it contains the densest city in the U.S. According to the United Nation’s 2011 Revision of World Urbanization Population Prospects, more than half of the world’s population lives in urban areas and is projected to increase to 67 percent by 2050 (2012). This increase is attributed to both the absorption of population growth in urban areas and rural urban migration. This projection shows the significance of studying a location with high density in its urban core. With the results of this paper, we may be able to apply this analysis to low income communities across the country to address the well documented underdevelopment in low income areas that plague the United States. Chapter One The Need for Intervention Many economic models show the importance of capital investment to productivity. The Cobb- Douglas function states productivity is a function of three factors: labor input, capital input, and a total productivity factor. As a result, capital is one of the components needed for productivity to occur. This function can also be applied to low-income neighborhoods. With restrictive access to credit, obtaining the capital needed to turn labor times the total productivity factor into output is restricted. Unfortunately, we see that in low-income areas access to capital is restricted and stunts growth. Malpezzi (2003) clarifies the function finance plays in the economy by separating the economy into two types of assets, real and financial. Real assets are defined as resources made to use other things, like machinery, labor, land, and knowledge. Financial assets are claims on the output of real assets. The significance of financial assets extends to the financial market as it itself is productive. Financial markets both public and private create efficiency through the matching of investors and savers. Without financial markets, those with investment opportunities must also be savers. Because of this, investments are postponed until the investor has enough money saved for the opportunity. Savers on the other hand must have knowledge of investment opportunities, which bolsters the view that savers and investors must be the same individuals. The lack of financial markets leads to the high transactions as participants must both save and find opportunities, so there is division of labor. Because of a lack of access to financial markets in 14 low-income neighborhoods, the same negative effects resulting from an absence of a financial market is seen. According to an Institute for a Competitive Inner City study of Community Reinvestment Act (CRA) data, the lowest income census tracts within inner cities get approximately 79% of the loans they would expect to obtain based on the number of firms located in those areas (ICIC, 2011). The CRA is a national policy that uses a similar model to address the same issues as the NMTC Program. It is somewhat of a precursor to the NMTC Program. Former Federal Reserve Board Chairman Ben Bernanke discussed the need for programs like the CRA in his speech at the Community Affairs Research Conference in 2007 as tools to revitalize urban decay in low- income and minority neighborhoods. He cited that many believe the reason for the decay was limited availability of credit. Limited access to credit is attributed to multiple factors; the first being racial discrimination. Bernanke (2007) states, “… racial discrimination in lending undoubtedly adversely affected local communities.” This discrimination dates back to 1935 when the Federal Home Loan Bank Board had the Home Owners' Loan Corporation create a level of security or risk map for real estate investments throughout cities across the country. The riskiest types of areas were coded in red and contained neighborhoods with a high percentage of African Americans. Private lenders followed suit with similar maps, which led to the term red lining. The 1961 Report on Housing by the U.S. Commission on Civil Rights stated larger down payments were required and blanket refusals were given to those who lived in the red zones. This became a significant obstacle for minorities to access credit. Other obstacles in accessing credit are related to information issues. Credit evaluations are more expensive as their credit histories are likely shorter with gaps. A lack of wealth accumulation and collateral makes it even more difficult to obtain credit as well. Without detailed information, like that displayed in a long credit history report, it is difficult for lending institutions to decipher which individuals and small businesses would default on their loan. The significant costs associated with collecting credit related information and trouble of keeping the information proprietary deterred lending institutions from entering the low-income market. Daniels and the Economic Innovation International (2005) categorize Bernanke’s reasons into five specific categories for why there is difficulty of obtaining capital in low-income communities. These include: (1) inadequate risk 15 pricing and pooling, (2) huge transaction costs, (3) inadequate market competition, (4) market prejudice, and (5) market distortion through government policies. Unfortunately, the Great Recession from 2007-2009 has exacerbated the difficulty to obtaining capital in low-income communities. The Federal Reserve of San Francisco’s 2010 working paper on CRA and small business lending in low and moderate income neighborhoods stated the number of small business loans dropped from 5.2 million in 2007 to 1.6 million in 2009. The amount of credit available to small business fell from $137 billion to $73 million in the same time frame. In order to manage the impact of the rapid economic change on neighborhood economies, the Organization for Economic Corporation and Development (OECD) has promoted the idea that national economic policies include both regional and local economic development in their plan. The OECD details goals of regional and local economic development, which include job creation, increase long-term career options for local residents, the inclusion of the disadvantaged and minority groups in the local economy, and improving local institutional and physical infrastructure to encourage a better business climate. While many variables lead to economic development, a common factor needed to achieve these goals is finance. The NMTC Program is a national economic policy that addresses both regional and economic development by financing projects within low-income neighborhoods. The following puts the Program into context by showing the evolution of national economic development policy from the 20 th century on. Overview of the Local Economic Development The general definition of economic development is increasing well being or welfare, but what does that really mean and how do we achieve this? Economic development in the first half of the 20 th century was defined as efforts to spur economic growth, including efforts to increase employment or create wealth. Blakely and Leigh (2010) build upon this definition to consider distributional considerations and environmental issues we are seeing today. They offer a three part definition: (1) “…economic development establishes a minimum standard of living for all and increase the standard over time (2) economic development reduces inequality (3) economic development promotes and encourages sustainable resource use and production” (pp. 75). The reason environmental considerations within the third part of the definition were added to 16 minimize the inequality in living standards between current and future generations. Like in many fields, there is not one theory or method to addressing the maintenance and improvement of living standard as well as reducing inequality in an environmentally conscious manner. The four theories we will focus on are the Neoclassical Economic, Economic Base, Product Cycle, and Location Theories. The Neoclassical Economic Theory assumes: individuals have rational preferences among outcomes; firms will maximize profits while individuals will maximize utility; individuals will behave independently with perfect information; and there is mobility of capital. This theory states that economic systems will reach a natural equilibrium given the assumptions. The Neoclassical Economic theorists are in favor of deregulation of markets. The idea is that without government intervention and regulation, economic systems will reach a state of equilibrium. This would mean that capital would flow from areas with high wage and property values to low- income neighborhoods to take advantage of lower labor and land costs. In cases of unemployment, neoclassical theory would suggest laborers to move to new employment areas. There are many critiques of this theory and examples of it have caused extreme income inequality. The first takeaway is that communities must use what they have to attract capital. Also, barriers to economic development include government bureaucracy and a poor business environment. Economic Base Theory proposes that a neighborhood’s growth is strongly associated with the outside demand for its goods and services. This theory divides a neighborhood’s economy into two sectors, a basic and non-basic. The basic sector consists of local businesses that rely on customers from outside that neighborhood. An example of a strictly basic industry would be mining, where the customers are likely outside of the neighborhood’s households. The non-basic sector includes businesses that offer products or services consumed by those within the neighborhood. This would include conveniences stores and dry cleaners. The Economic Base Theory would lead to economic development policies that attract or support businesses that cater to international or national markets. This theory is more short-term in nature as it focuses on attracting businesses to add to its current economy, which will change over time. One of the disadvantages of following this theory is that it ignores internal customers’ needs. Also, 17 companies that cater to external or international customers tend to be larger or have parent companies located outside of the neighborhood. This leads to a leakage of the income, so instead of income being spent or recycled within the neighborhood, it is spent elsewhere and lost. The Product Cycle Theory links the innovation and dissemination of a product to an area’s economy. In this theory there are three stages in the product cycle; new product stage, maturing product stage, and the standardized product stage. This theory proposes that that the research and development of a product occurs within higher income areas as capital is required to fund this process in the first stage. Areas with higher incomes are also needed to purchase the product before the product becomes standardized. As the product moves into the maturing stage, it becomes standardized and sold in the mass market. As the product and production process become standardized, lower income areas can usually reproduce the product at a lower cost. In this stage the production no longer needs specialized workers or firms, which is the standardized product stage. This theory tells more about the place of existing firms within an area. Location Theory started with examining where firms decided to locate based on the process of their product and transportation costs. If a firm’s product got larger and heavier through the production process, they would locate closer to their customer base to minimize transportation costs. As technology advanced, the theory encompassed more considerations firms have when deciding where to locate, such as warehousing needs, labors costs, communication, energy costs, location of suppliers, and more. The concept of economies of scale and agglomeration economies falls under this theory. Firms will benefit if they locate in close proximity to their suppliers and other similar firms. Economic development policies that follow this theory usually focus on one of these considerations either through incentives or subsidies to create an advantage they would not normally have. These theories have resulted in two main models. The first is the attraction model. The various forms of the attraction model entice both firms and individuals to locate within the specific area by providing subsidies and other incentives. By providing incentives and subsidies to firms, the desired outcome is that these firms will generate economic growth and wealth greater than the amount of public funds spent to attracting the firms. Incentives are also provided at the 18 community level to attract a specific type of population group. For example, many cities have attempted to attract young, middle-class to upper middle class individuals who have a higher income to spend in the city and also are associated with a higher education attainment or experienced skill set that will in turn attract firms to the area. In order to attract these individuals, cities invest in creating an atmosphere they assume this population will prefer, such as communities with more independent coffee shops. In this case, the city or neighborhood is packaged and marketed like a product. The second model is the New Markets Model, which directs funding and development to areas that are being underutilized. Policies following this model dispense aid to specific geographic areas to spur economic redevelopment. They focus on growing economies of areas that have seen a decline, especially inner cities. Michael E. Porter brought attention to inner cities, when he wrote the “The Competitive Advantage of the Inner City.” He proposed a paradigm shift in the way we think about inner cities in that we should think of them as the center for opportunities rather than a drain on the city. Inner cities are currently being underserved, which presents a potential market as they are characterized by an extremely dense population with a sizable collective purchasing power. New York Governor Andrew Cuomo extended the same idea of underserved communities with large purchasing power to rural areas with new migrants. One of the key components to the New Markets Model is that subsidized investment is needed to divert investment into these areas to compensate for risk. The Evolution of National Economic Development Policy We can see these theories and models applied throughout the history of national policy within the United States. These type of policies date back to the Louisiana Purchase in 1803 as the purchase secured Americans the right to trade in New Orleans (Blakely & Leigh, 2010). The Morill Acts of 1862 and 1890 incentivized human capital development by funding the land-grant university (Blakely & Leigh, 2010). The next major economic development program was the New Deal that addressed economic needs across the entire country after the Great Depression. It is not until the 1937 Housing Act that we see the first national policy with an economic development component directed at low-income communities. This Act provided affordable housing assistance along with employment opportunities and economic stimulation. The Urban 19 Renewal Program, part of the 1949 Housing Act, was the first time economic development was one of the main objectives of a national policy. Rural areas were not forgotten as several rural development programs were created during the same time, such as the Rural Electrification Act of 1935. In the 1930s and 40s, government programs were geared towards creating and supporting businesses through land development, loan packages, tax abatements, and loans. The programs were based on firm behavior and how public efforts could affect the decision process of where industrial firms would locate. In the 1950s President Eisenhower approved the Small Business Act that created the Small Business Administration, a federal agency that aids and assists entrepreneurs and small businesses. This act also approved the Small Business Services Guaranty, which addressed the financial needs of small businesses that were unable to get loans without the assistance. We see the same trend with the Small Business Investment Company Act of 1958 that created the Small Business Investment Company (SBIC) Program. This program allows the Small Business Administration to license privately owned investment firms. These private investment firms then use a combination of their own capital with low rate federal loans to invest in entrepreneurs. In the 1960s we see two shifts in the direction of national policy. First, we a shift from policies directed at small businesses to policies directly geared towards employment and training. The Public Works and Economic Development Act of 1965 provided “grants for public works and development facilities, other financial assistance and the planning and coordination needed to alleviate conditions of substantial and persistent unemployment and underemployment in economically distressed areas and regions” (Blakey & Leigh, 2010). The second shift is the policy to extend those successful initiatives from the New Deal to inner cities, with the Model Cities and New Communities Programs. The focus on urban areas continued through the 1970s, the Community Development Block Grant of and Urban Development Action Grant provided funding to local governments to provide housing and employment opportunities to low income individuals. The Community Reinvestment Act (CRA) of 1977 was directed at low income communities but in a different capacity. The CRA was designed to reduce redlining and encourage banking institutions to serve the needs of all members of their community, including low and moderate income residents. In order to achieve this, all banking institutions that are insured by the Federal Deposit Insurance Corporation are to be evaluated on whether or not they operate in all areas they are charted to. In the 1980s, there was a decrease in urban development 20 programs. In addition, the New Communities Program and Urban Development Action Program were discontinued. During this time the Low Income Housing Tax Credit Program was enacted to provide tax credits to real estate developers to attract investors in projects that increase the amount of affordable housing. Even though this project does not focus on the urban core, it does provide historical context for the NMTC Program as the structure of both programs are similar. We see the return of urban development policies in the 1990s as it became apparent that the challenges of cities and suburbs were connected to inner cities. We also see the addition of market based strategies to address economic development. During this time we see a trend of various location based economic development policies such as Empowerment Zone Initiative, Neighborhood Revitalization Zones, HUB zones, and the Gulf Opportunity Zone in 1993. These policies strictly follow the New Markets Model. The New Markets Model is part of the general conversation of how we should approach economic development today. Blakely and Leigh describe the main focus of the economic development discussion since the 1990s has been centered on three main approaches. The first approach utilizes tax incentives and federal funds to rebuild industrial infrastructure throughout the country. The second approach is to decrease government regulation similar to that of the Neoclassical Theory. An example of this approach would be less regulation in the labor market, such as eliminating or floating the minimum wage. This would theoretically allow unemployment to reach its natural state and absorb more of the unemployed that are willing to work into the labor market. If the minimum wage is greater than the equilibrium wage, then it will lead to unemployment because the quantity of labor demanded will be less than the quantity supplied. The third approach says that economic development should focus on firms alone but include the local community. This would include national policy that gives local communities greater input in and control over the type of corporate investment policies and their quality of life. What sets the NMTC Program apart is that it is a combination of multiple approaches to address the restrictive capital market in disadvantaged areas and compensate for the asymmetrical information market failure. Not only does the Program utilize a combination of the Blakely and Leigh’s third approach and location based incentives, it looks at past federal economic development programs to address the problems of disadvantaged communities, specifically the lack of access to capital. 21 Overview of the New Market Tax Credit Program The NMTC Program was devised using the New Markets Model, which focuses on reaching markets that have high purchasing power potential but have yet to be tapped in to. The U.S. Department of Housing and Urban Development (HUD) stated that inner-city neighborhoods in America had a potential retail purchasing power of $331 million in 2000. The reason these inner-city neighborhoods have been underserved is due to the asymmetric information and beliefs that income levels were too low and the crime rate were too high within these areas. In actuality, crime rates have decreased to the lowest in the past 30 years due to a wealthy national economy. Inner city populations enjoy high pedestrian traffic and strong purchasing power due to population density. While economic various development strategies may increase new investment or reinvestment in these economic opportunity zones, Blakely and Leigh (2010) express that some form of subsidy is needed for progress to occur. This is where the NMTC Program comes in. The NMTC Program incentivizes investment into inner-city and low-income communities through tax credits. The NMTC Program allocates tax credits to certified Community Development Enterprises (CDEs), who then attract private investors for projects within low- income communities with these credits. This allows the federal government to rely on the expertise of nonprofits and for profit entities. CDEs must have the following three characteristics to qualify for the program. First, the CDEs must be a corporation or partnership within the United States. Second, CDEs must demonstrate a history of serving or providing capital to low- income communities. Third, CDEs must have a resident of the low income community on their advisory board to represent the interested of the area and for accountability measures (CDFI, 2013). Once the CDE is certified by the CDFI, they can then apply for NMTCs. Applications are then reviewed by the CDFI Fund on an annual basis. Certified CDEs can send their application to the CDFI Fund for an allocation of credit. These applications are then scored in four categories. The first category is community impact, which is the extent to which the CDE impacts the community. This includes the extent to which community representatives are involved in the decision making process and whether or not the CDE can demonstrate its impact 22 on the community. The second category is the strength of CDE’s strategy to identify and address the needs of the community through financial products. The third category is the CDE’s capitalization strategy, which includes the CDE’s plan to attract investors and secure funds. The last category is management capacity, which scores the ability of the CDE to carry out an effective NMTC strategy. This includes but is not limited to the CDE’s experience, staff, and partners. Once applications are scored, the CDFI Fund and selected CDEs go into an agreement that allows CDEs the right to market tax credits and carry out their strategy detailed in their NMTC application. The CDEs then attract taxpayers to secure investments, also known as qualified equity investments (QEIs). QEIs consist of stock options or capital in the CDE. In exchange for their investments, taxpayers receive federal income tax credit equal to 39% of their initial investment over a 7 year period. A credit equal to 5% of the QEI is taken in the first three years and is increased to 6% for the last four. In addition to the tax credits, investors receive a return on investment (ROI) like they would on any other investment. The CDE then invests in qualified low-income community investments (QLICIs). Figure 1 illustrates this process. Figure 1- New Market Tax Credit Capital Flow Chart (2) Invests in (1) Allocates tax credits to (5) Claims tax credits (3) Invests capital & return on investment & provides financial services or counseling (4) Return on Investment Community Development Financial Institutions Fund (CDFI) Tax Paying Investor Community Development Entity (CDE) Qualified Low-Income Community Investment (QLICI) 23 The QLICIs are then invested in Qualified Low Income Community Businesses, which are businesses that have strong ties to a LIC. Strong ties are defined as having 50 percent or more of a business’ gross income comes from operation within an LIC, at least 40% of the use of the tangible property is within an LIC, or 40 percent or more of the services rendered by the business’ employees are completed within an LIC. Although the types of qualified businesses are vast, there are some restrictions. These restrictions include the operation of residential rental property, properties where no major or significant improvements have been made, the development or holding of intangibles for sale or license, operating gambling facilities, operating golf clubs, or operating an liquor retail shop. These restrictions were created to increase the quality of life for local residents and limit redundancy between programs. The Low Income Housing Tax Credit Program is similar to the NMTC Program, but it provides credits for residential rental properties. These restrictions along with other program requirements have led to the majority of NMTCs funding real estate projects, where the additional investments went to fund gap financing. Gap financing is the difference between the total cost of the project and the debt and equity raised. This trend is consistent with funded projects within New York. CDE representatives reported that the term loans were where CDEs made most investment in Qualified Active Low Income Community Businesses (QALICBs). The GAO found that from 2003 to 2008, over 85 percent of total NMTC dollars dispensed by CDEs consisted of term loans. CDEs provide more favorable financing to QALICBs by either providing lower than market rate interest on their loans or to leverage additional investment dollars that the QALICBs already have in place. Figure 2 below provides an example of a leveraged NMTC transaction, where the tax credit investor forms a limited liability entity that obtains a loan from a bank. The additional funds from a leveraged lender (i.e. bank) are required by program regulations. The investor then combines funds and invests in a CDE, who then invests them in a QALICB. The funds are then separated out into two loans for tax purposes. The leveraged lender receives interest payments on the loan by the limited liability entity and the tax credit investor receives a return on their investment from the tax credits. 24 Figure 2 Example of Leveraged Loan Transaction Loan 1 interest rate Loan 2 Interest Rate + Tax credits Makes a Qualified Equity Investment (QEI) of $1,000,000 Makes a $ 950,000 qualified low income community investment (QLICI) Tax Credit Investor Investor claims up to $390,000 in tax credits (1,000,000*.39) Credit Price = 75 cents Tax credit equity to Investment fund = $292,500 ($390,000*$0.75) ABC Bank $707,500 NMTC non-recourse loan @ a below market interest rate Investor limited liability entity Total fund: $1,000,000 (Tax credit equity $292, 500 + leveraged loan: $707,500) Community Development Entity (CDE) 5% fee = $50,000 ($1,000,000 * .05) Qualified active low-income community business (QALICB) Loan 2: Levered Debt $682,500 @ below interest rate The CDE provides a second loan to the QALICB, with a below market interest rate, on which the QALICB will make payments during the 7-year NMTC compliance period. At the end of the 7-year period, the QALICB will need to refinance this loan. Loan 1: Tax Credit Equity $267,500 @ 1% interest rate The CDE uses funds from the sale of tax credits and loans at a low interest rate with the intention of converting the debt to equity in the QALICB following the 7-year period in which the investor can claim the NMTCs 25 Previous Literature on NMTC Program This thesis paper hopes to add to the existing discussion on the NMTC Program. The following is a representative of the analysis that has been completed on the NMTC Program, including both governmental and academic work. The Urban Institute (UI) analyzed a diverse group of five projects resulting from 2002 and 2003 NMTC allocations. These projects were chosen to represent the diversity in projects in three categories: location, participants, and project type. The data used in this analysis comes from CDEs themselves and includes the financial details. This report had two goals. The first goal was to provide a complete picture and more detail on the selected projects than previously provided. The second goal was to use the strength of the case study approach in hopes to design a more inclusive evaluation method. The Urban Institute gathered a significant amount of detailed information at the project level and examined the information from a variety of perspectives. UI (2007) acknowledged the limitations of this method as (1) the early initiation of the projects may not represent all projects fully (2) preliminary research did not provide the ability to complete a detailed review of business plans, financial projections, or content with those outside of primary stakeholders, (3) not observing the projects first-hand. The findings of this study were focused on finding trends across the five projects. They found that the projects examined had varying levels of involvement by local government and planning agencies. 4 out of 5 projects were real estate related. This is consistent with findings by the GAO (2009) and San Francisco Federal Reserve Bank. While none of the projects defaulted, 2 projects had loan repayment issues in the beginning. All projects were completed on schedule. 3 out of 5 projects evolution, referring to how the various stakeholders became involved, had either “either the existing businesses or some of their principles had previous customer relationships with either the CDE or investors” (Abravanel, Pindus, & Theodus, 2007 pp.12). In all cases, available financing was either insufficient or not available on more favorable terms. While all projects received financing, none received financial counseling. The UI found through interviews and documentation that NMTCs were not always necessary. UI noted that CDEs had sizable variation performance criteria. The New Markets Tax Credit Coalition also relies on the CDE feedback for analysis purposes. They complete an annual survey of CDEs who received NMYC allocations. In 2013, they received feedback from 72 CDEs, which represents 40% of total allocated funds for the year. 26 The goal of this survey is to gather information on NMTC activity and provide insight to trends within the surveyed year and compares it to previous years. In their 2013 report, the NMTC Coalition found that demand for tax credits remained strong, by saying 314 CDEs requested allocation equivalent to $21.9 billion with only 85 of them were awarded $3.5 billion in total. 70% of CDEs use NMTC capital within 1 week, which has increased 5% over the last 2 years. This is the first year that the NMTC Coalition has asked CDEs to report the number of jobs created. 20,251 full-time jobs and 27,570 construction jobs were created in 2012. Full-time jobs are defined by permanent jobs either contributing to business operations or construction jobs are those pertaining to the construction or renovation of infrastructure or land. Over half of the full- time of the jobs created in 2012 were within three sectors: industrial or manufacturing, and healthcare & community facilities. Trends in 75% of the projects were located within severely distressed communities, which included having one or more of the following characteristics: poverty rate greater than 30%; and a median income less than the area median income, or an unemployment rate equal to at least 1.5 of the national average. 65% of jobs were created in communities that had an unemployment rate of at least 1.5 times the national average. In 2012, there was $55 billion in total investment of which $27 billion was NMTC related capital. Based on the Joint Committee on Taxation estimates, the NMTC Program cost the federal government $7 billion in revenue, which results from an effective rate of revenue loss and cost to the government of 26%. The NMTC Coalition also completed an economic impact report for more detailed assessment of the NMTC Program’s impact. More than 2,900 projects were examined by their IMPLAN model. IMPLAN is an economic assessment software system that analyzes a combination of location specific trade patterns, demographic statistics, and other economic variables. The IMPLAN model provided information on both jobs and tax revenue for daily operating activities as well as construction activities. Because constructions jobs are temporary by nature, the number of construction jobs is reported in years of labor for one person. For example, 20 construction jobs can be 80 workers on 3 months contracts or 4 people working for 5 years each. Between 2003 and 2010, 535, 874 jobs were created; 200,537 operational and 335,337 construction jobs. While the following estimates are composed of 2003-2010 tax revenue for construction related activities, only 2010 operational tax revenue was available. The NMTC Program resulted in federal tax revenue of $5,386,807,565 and state and local tax revenue of $3,146,449,690. IMPLAN was also able to break down the total number of jobs 27 created into those directly, indirectly, and induced by NMTC investments. 296,577 jobs were directly linked to NMTC investments, which includes construction jobs to create facilities and permanent jobs relating to daily operations within that industry. Approximately two-thirds of the directly linked jobs were construction based. Jobs created indirectly encompass those resulting from intra-industry impacts. A total of 96,880 jobs were created within 2003-2010 with over two-thirds related to construction activities. Lastly, induced economic effects include the increase in household spending as result in an increase in consumer demand. For example, if the NMTC investment funds a manufacturing plant, additional retail may be built to service the increase in demand of those plant workers. 142,417 jobs were induced by the program with almost 70 percent construction related. The U.S. Government Accountability Office (GAO) has also looked at the NMTC Program as it is their responsibility to examine how the federal government spends public funds. Within their 2010 report on the program, they analyzed data from the CDFI Fund's Community Investment Impact System (CIIS) data, interviews with CDEs, and looked at nine case studies of CDE's to analyze the trends within the Program. The U.S. GAO found that the majority of projects funded were related to commercial real estate and used for gap financing. This may be due to nature of real estate projects as being fixed to a specific location and long-term projects, which makes them easier to comply with the NMTC Program requirements. The GAO Report also stated 90 percent of the projects funded were within metropolitan areas, which is something that the CDFI would adjust in the future to increase the amount of funding going to rural areas. In terms of the impact on low-income neighborhoods, GAO stated part of the challenge in the finding the extent to which the NMTC Program has benefitted a community is the size of funded projects compared to the total economic activity within that community. It is likely the size of the projects is much smaller, which makes it difficult to distinguish effects attributed solely to the Program. In addition, the GAO reported the difficulty of testing whether investment would have occurred without the program is to determine what investors and developers would have done without the Program. In 2007, the GAO used a fixed effects model to compare the level of assets and the growth of assets between investors that participated in the NMTC Program and those that did not. The model is Yit = Xitβ + μit, where Yit represents the log of total assets, total liabilities, or net assets; Xit represents control variables which include the lag of net assets, the 28 NMTC participation dummy, year dummies, and region dummies; β represents the estimator of the relationship between the control variables and the log of total assets and μit represents a random error term. Investors who participated in the program were found to have higher net assets; however, the results were not robust over different specifications of the model. Growth rates of net assets were not statistically significant different between groups, which means that both groups of investors were investing at similar rates. The analysis did find that corporate investors were likely to shift their investment away from higher-income communities to those that qualified for the program rather than increase their investment budget in low-income communities. Most recently the question has moved to whether the program is worth it. While we have self- reported job counts from surveys and the IMPLAN model, the next analysis takes the discussion of the program’s efficacy one step further. Freedman (2012) uses an impact analysis to examine the effects of NMTC Program on poverty levels and land value using a quasi-experimental method. Freedman used the discontinuity in the way census tracts can qualify as for tax credits in order to create a pseudo-random assignment. Census tracts below the income threshold qualify as low-income and are eligible for tax credit allocations, where those above the threshold do not. Outside of this, the census tracts on either side of the threshold are similar. The study hopes to gain insight to casual inferences of the effects of the NMTC investment by comparing census tracts with a narrow range centered on the threshold. The aim of the study is to solve for β 1 in the following equation, . Here Δy i represents the change in either poverty rate or median home value in individual census tracts. More than one independent variable was examined in this study. N i represents the amount of NMTC subsidized investment in an individual census tract and X i is a set of baseline characteristics, which were obtained from the 2000 Decennial Census. He was unable to control for the decision making factors of which tract a CDE chooses to allocate funds. For example, a CDE is likely to base their decision on both the current and future characteristics of each tract, which may not be entirely captured in X i and affect Δy i. This can lead to an endogeneity problem, where the error term ε i to be correlated to N i and bias the estimate of β 1. To address of issue of endogeneity as well as exploit the discontinuity at the income threshold, the author uses a regression discontinuity (RD) model. The author uses the ratio of the tract MFI to the state’s of MSA’s MFI to define the threshold. Tracts with a ratio 29 less than or equal to .8 can be assigned eligible for NMTCs while those with a MFI greater than .8 do not. While the majority of tracts qualify for the program, there are a few tracts that qualify for the program based on other variables for the actual NMTC program. The MFI ratio is then used for the range of census tracts to look at, which is 0.7 to 0.9. Low income community designation is then assigned randomly to tracts within this range. Because low income designation affects other outcomes only through its effect on the location of where investment will occur, it can be used as an instrument variable for investment. The first-stage regression equation is N i = α 0 + α 1 LIC i + f (m i ) + X i Σ + V i , where N i is defined as investment like that in the previous equation. LIC i is a treatment variable that is defined as 1 when the tract is a LIC and 0 if not. m i is a forcing variable that denotes the MFI variable. Freedman uses the following form for specifications for the control function, where p is the order of the polynomial. The reduced form of the LIC designation on the neighborhood outcomes is . The IV estimate of β 1 is the ratio of γ 1 and α 1. Using the OLS regression equation 1, this study finds that the subsidized investment has only a slight positive effect on the neighborhood variables in low income communities. An increase of $1 million of NMTC investment is associated with an increase of 0.02% change in median household income keeping all other variables constant. Also, an increase of $1 million of NMTC investment is associated with a decrease of 0.01 -0.03% in median home values. The OLS regression also results in negative relationship between investment and poverty and unemployment rates. However, these relationships are statistically insignificant. Also, “the results suggest that some of the observed impacts on neighborhoods are attributed to a change in the composition of residents rather than improvements in the welfare of existing residents.” While Freedman has approached this topic using an econometric model; this paper will offer an alternative model in Chapter Three. 30 Chapter Two - Case Studies of Projects The following is a case study of 5 projects located within the State of New York. They were chosen based on the success of the associated CDE. Because CDEs need to prove that they have used previous allocations appropriately, those CDEs with repeated allocations have been deemed successful in use of NMTCs. The reason both case study and empirical approaches are used in this paper is to emphasize the fact that a more comprehensive approach needs to be created in order to evaluate the impacts of this program. The method used to analyze the projects is based on that used by the Urban Institute. Additional characteristics were added for comparison purposes between approaches. Also, more recent projects were included in the case study approach to ascertain trends across the life of the project rather than a specific point in time. The goal is to zoom in on specific projects within New York using the existing framework for how impacts are analyzed. HEDC New Markets Inc. is a subsidiary of the National Development Council (NDC), one of the oldest non-profit community organizations in the United States, founded in 1969. Because only for-profit CDEs can accept quality equity investment from NMTC investors, a for-profit entity was created. National Development Council’s mission is to increase the flow of capital to underserved areas to create jobs and community development. They offer professional training and assistance to community development partners, small business financing, as well as debt and equity for various real estate projects. NDC is a national organization with client relationships in over 40 states and Puerto Rico. They have extensive experience with other economic development programs, such as Low Income Housing Tax Credits (LIHTCs), Rehabilitation Tax Credits (RTCs) for historic properties, U.S. Department of Housing and Urban Development (HUD) Section 108 loans, SBA guaranteed financing, tax exempt debt, and HUD’s Urban Development Action Grants (UDAGs). Like NDC, HEDC New Markets Inc. services both urban and rural areas across the country. Even though they service the entire nation, their initial application for a NMTC allocation focused on projects within California, Maryland, New York, Washington, Illinois, Indiana, and Pennsylvania. HEDC New Markets Inc. is frequently allocated New Market Tax Credits and has won credits in 8 rounds of allocations worth $704 million. Of this, $639 million have been 31 invested in 78 projects, which have brought in $1.4 billion in total investment. Through 2010 HEDC New Markets Inc. has been involved in 9 projects within New York in both the city and surrounding areas. Local Initiatives Support Corporation (LISC) is another national non-profit organization focused on providing disadvantaged communities with access to public and private resources. LISC was established in 1979 by the Ford Foundation. Since 1980, LISC has leveraged $12.9 billion for $38.3 billion in total development. They have 30 offices throughout the country. LISC has been allocated $778 million worth of tax credits in 8 rounds of NMTCs. To date, $600 million has been invested to create over 17,000 temporary and permanent jobs. LISC focuses on using NMTCs on real estate developments, which is both a factor of the nature of the program and LISC’s core competency in real estate. LISC has been involved with 8 real estate projects within New York. More specifically, LISC uses NMTC primarily for debt financing. LISC’s NMTC activity is managed through New Markets Support Company (NMSC), its tax-credit equity affiliate. NMSC is a nonprofit based out of Chicago that attracts investment to commercial real estate and operating businesses to revitalize neighborhoods. NMSC has involved in a variety of projects in low income neighborhoods including major retail developments, manufacturing/industrial sites, charter schools, theaters and urban entertainment districts, athletic facilities/fields, office space and health care centers. Chase New Markets Corporation is a subsidiary of J.P. Morgan and Chase Company. Chase participates in three ways in respect to the NMTC Program. First, Chase acts as an investor with $1.4 billion allocated to outside CDEs. Second, Chase New Markets Corporation is Chase’s CDE with $405 million in NMTC allocations. Lastly, Chase provides loans for various NMTC projects. Chase New Markets Corporation is part of Chase’s Community Development Banking Group that provides below market interest rates on senior debt to CDEs, CDFIs, not-for-profits, real estate developers and operating businesses. Chase New Markets focuses on commercial and nonprofit projects to offer loans at below-market interest rates and equity-like financing. Chase New Markets has been allocated $405 million in tax credits from 6 rounds. They had utilized these credits to support 3 projects within New York. 32 Empowerment Reinvestment Fund is the CDE subsidiary of TruFund Financial Services headquartered in New York. TruFund Financial Services is a national non-profit whose mission is to spur economic development by providing capital and business assistance to small businesses, nonprofits, and real estate developers in underserved communities. TruFund Financial is the new brand name of Seedco Financial, who updated and refined their image and strategy in conjunction with the name change. They focus on servicing clients located in Alabama, Louisiana, and New York. TruFund Financial Services works with community based partners to address lack of capital in local markets, which can range from small microenterprise loans to local businesses to large projects requiring millions in NMTCs. To date, the Empowerment Reinvestment Fund has received $182 million in credit through 6 allocation rounds and has been involved in 4 projects within New York. Nonprofit Finance Fund (NFF) was founded in 1980. NFF is a national, nonprofit Community Development Financial Institution (CDFI) founded in 1980. They primarily service nonprofits or organizations that are driven by mission rather than profit. In 2000, NFF widened their focus to include the capitalization needs of nonprofits compared to before where they concentrated more on facilities. Compared to the other CDEs examined, NFF is a later comer with their first allocation round in 2006. Since 2006, they received allocations in 6 rounds worth $ 231 million. Table 1 below illustrates the similarities between the most successful CDEs in New York. 33 Table 1 Characteristics of Selected CDEs CDE National Development Council, HEDC New Markets Inc. Local Initiatives Support Corporation Chase New Markets Corporation Empowerment Reinvestment Fund Nonprofit Finance Fund Location New York, NY New York, NY New York, NY New York, NY New York, NY Allocation years and amounts 2002: $30,000,000 2003: $135,000,000 2006: $121,000,000 2008: $90,000,000 2009: $110,000,000 2010: $63,000,000 2011: $90,000,000 2012: $65,000,000 2002: $65,000,000 2005: $90,000,000 2006:$140,000,000 2007: $133,000,000 2008: $80,000,000 2009: $115,000,000 2010: $70,000,000 2011: $85,000,000 2006: $50,000,000 2007: $60,000,00 2008: $85,000,000 2009: $40,000,000 2011: $100,000,000 2012: $70,000,000 2002: $10,000,000 2003: $25,000,000 2006: $40,000,000 2008: $35,000,000 2010: $35,000,000 2011: $40,000,000 2006: $20,000,000 2008: $50,000,000 2009: $60,000,000 2010: $21,000,000 2011: $40,000,000 2012: $40,000,000 Geographic scope National National National National National Organization Structure Non-profit; for profit subsidiaries created for individual projects Non-profit; for profit subsidiaries created for individual projects For-profit Non-profit; for profit subsidiaries created for individual projects Non-profit; for profit subsidiaries created for individual projects Controlling Entity Structure National non-profit National non-profit For-profit National non-profit National non-profit Intended Use of NMTCs to provide operating business financing to offer more favorable debt and equity products to community development projects within its program areas to invest nationally in operating businesses, charter schools, health care facilities, grocery stores, and community centers to offer flexible loan products affordable financing and technical assistance to small and mid-sized non- profits From Table 1 we can see the characteristics that most successful CDEs in New York exhibit. All 5 CDEs are large organizations that service clients throughout the country.4 out of 5 are nonprofits that have a history and mission aligned with the goals of the NMTC program. While Chase New Markets Corporation is the only subsidiary of a for profit organization, the Program is aligned with its parent company’s commitment to bring $800 billion to make loans and investments in low-income neighborhoods. This commitment was made in 2004, which is likely why the first year Chase New Markets Corporation received NMTCs was 2006, 4 years later than 3 of the other CDEs. The intended use of the tax credits are consistent among these CDEs in that the tax credits would be used to offer debt and equity to operating businesses. The only CDE to stand out in its intentions is the Nonprofit Finance Fund that offers technical assistance in addition to financing products. Also, NFF explicitly targets “mission driven” organizations, while the other CDEs do not specify what type of organizations they work with outside of those who qualify for the NMTC program. From literature provided, the other CDEs utilize their 34 NMTC allocations based on the type of project rather than the type of organization or business leading the project. Although characteristics of CDEs may lead to insight on which projects they are involved with are successful in terms of being completed, we are more interested in to what extent do the communities in the project area benefit. The projects examined represent the diversity and trends of projects these successful CDEs are involved with. PROJECTS HEDC New Markets Inc. used its 2008 allocation to revitalize the historic Yonkers Recreation Pier. This project consisted of the restoration of the 20,000 square foot Pier. This two story mixed use project is part of an ongoing economic development push in Yonkers that has occurred over the past 15 years. HEDC partnered with the City of Yonkers to carry out this project. Part of the City of Yonkers’ strategic short-term goals includes revitalization of the waterfront in order to attract job creating companies and retail. Yonkers attributes include the fourth largest city in New York based on population, with a poverty rate of 15.5%. The state of New York had a poverty rate of 11% in 2000; showing Yonkers is an area in need of assistance in respect to economic development. Their unemployment rate is 6.3 percent. In terms of financing, HEDC used $7.5 million to fund the $12 million renovation project. They partnered with Washington Mutual. This project has also utilized an additional $950,000 of federal funding through HUD Section 108 Loan Guarantee Program. Other sources of public funding for this project came from the Port Authority of NY & NJ funds and HUD Economic Development Initiative grant funds. As a result of the pier being restored, 96 jobs were created. Many of these jobs were filled by local residents. Job creation is one component of the positive impacts this project has had on Yonkers. Additional recreation space and public access to the Hudson River is the other. This project has also attracted Peter X. Kelly, a local and James Beard award winning chef, as a location for his restaurant X20 Xavier, which has bolstered economic development of the area by attracting press and visitors to make the Yonkers Pier a destination for both locals and tourists. The success of this project was recognized by the Novogradac Community Development Foundation as the best NMTC project in the operating business category for 2009. 35 LISC used its 2007 NMTC allocation to raise funds for the relocation of the Center for Urban Community Services headquarters from 120 Wall Street to East Harlem. East Harlem in located on the northeast side of Manhattan. It has a population of 117,743 and a poverty rate of 44 percent. The Center for Urban Community Services (CUCS) is a nonprofit human service agency with a mission to provide programs and access to services that assist homeless and disadvantaged individuals in overcoming issues of homelessness, unemployment, addictions, and other related issues. CUCS is an existing organization founded in 1979 as Columbia University Community Services. In 1993, Columbia University Community Services responded to the increase in the number of homeless individuals by increasing its geographical reach and its services to disadvantaged populations that were not catered to before. The NMTCs allowed CUCS to obtain and renovate a 25,200 square foot; six story commercial building into space for its administrative headquarters, Housing Resource Center, and Career Network. Before NMTC related funding, CUCS relied mostly on grants and private donations as a nonprofit organization and was having difficulty obtaining funds at an affordable rate. NMTCs provided a below market rate interest, a higher loan to value ratio, more flexible borrower credit requirements, and a longer interest only payment period. The relocation allowed many of its clients from the upper east side of Manhattan to access the CUCS’ services more easily. 91 construction jobs and 120 permanent jobs were created by this project. Because East Harlem has a high percentage of disadvantaged residents, it is likely that the greater access to CUCS services for many is more significant than the number of jobs created for this project. Chase New Markets Corporation (CNMC) used its 2010 NMTCs to fund an ongoing economic development plan to revitalize the New Scotland Avenue in Albany, New York. This economic development plan began in 2003 as the Park South Urban Renewal Plan with a goal to attract private investment, improve the quality of life for the neighborhood, and convert deteriorated or vacant properties into attractive space for retail and office space. In addition, another main strategy was to connect the primarily residential neighborhood through retail along the New Scotland Corridor. This included a surgeon pavilion for the Albany Medical Center to provide their surgeons with administrative and clinical office space. As a mixed use building, the first floor of the building is occupied by Panera Bread bakery and cafe. Many of CNMC’s use of tax credit allocations are used in raising funds in conjunction with another CDE’s allocation. For this 36 project, CNMC used $12,132,000 in tax credits, which is equivalent to 80 percent of the total development cost. Like the previous projects, the Surgeon’s Pavilion created both construction and permanent jobs; 135 and 250 respectively. Out of those permanent jobs 175 of them are full-time with the remaining permanent part-time jobs. This project also supported the ongoing economic development efforts within the New Scotland Avenue corridor with a Patient’s Pavilion opening up this year. In Brooklyn, ERF dispense credits to support the renovation of an old manufacturing building into a space used by light industrial producers and artisans. In 2007, Greenpoint Manufacturing and Design Center started to create an industrial center at 221 McKibbin Street, Brooklyn, New York. They are a nonprofit real estate development organization headquartered in Brooklyn whose goal is to plan, develop, and manage industrial properties to support the manufacturing sector. The organization began in 1992 and is a relatively small developer with 4 properties within Brooklyn. The McKibbin Street Industrial Center was a 72,000 square foot old manufacturing building that had deteriorated and most recently been used as a warehouse for imported furniture. In the 1980s to 1990s, artists began to move into North Brooklyn, which began the gentrification process of Williamsburg and Greenpoint. Shortly after, property values began to increase dramatically. Between 2000 and 2007, there was a 200-400% increase in industrial property values within the area (GDMC, 2010). In addition, rezoning in 2005 allowed residential use within the industrial waterfront properties exacerbating the problem of affordable rent for local businesses and artisans. This project was created to address the need to support artisans and existing local businesses by providing them a space to create at affordable rates. GDMC used funds from the New York City Council to acquire the property. Sovereign Bank and the New York Investment Fund provided the remaining financing for the project to begin in 2007. Unfortunately during the construction phase, GDMC encountered unexpected costs related to structural and foundation damage. In order to cover these additional costs, GDMC reverted back to their initial strategy of looking for tax credits to fund part of the project. Originally, it was difficult to obtain tax credits due to the nature of the project. Unlike with large companies or big box retailers, developers working with small or medium size businesses usually have issues 37 with the pre-leasing requirements and lack of credit tenants that are needed to qualify for QEIs. Eventually, GDMC was able to partner with Portland Family of Funds, a CDE, who brought in Citibank’s CDE who then brought in Seedco, the parent company of the Empowerment Reinvestment Fund. Through these partnerships, both New Market and Historic tax credits were eventually obtained 9 months into construction. ERF contributed $4.65 million in NMTCs out of the total $17.8 in tax credit allocations. It is very common for projects to be funded by multiple CDEs, especially with CDEs of large banking institutions, like Citibank and JP Morgan Chase. The financial structure became complicated with high transaction costs due to the addition of leveraged funds through the New Markets structure and the complex leasing structure used to accommodate the Historic tax credits. With the additional funds, GDMC was able to continue construction and complete the project in 2009. The project was able to restore the facade and structural damage as well as divide the property into smaller units. This project created 100 construction jobs. This project will create 100 permanent full-time jobs once completely leased out. The project created workspace for 15 small local businesses that face displacement due to real estate speculation, rezoning, and conversion on industrial properties to other uses. GDMC was also able to repair and repurpose a historic building within an area blighted by industrial properties. The Nonprofit Finance Fund used a portion of their 2009 NMTCs for the Pitkin Avenue Project, a mixed use project in Brooklyn, where the lower floors would be occupied by the retail and the upper floors by a charter school. The project involves repurposing the former Pitkin Loew’s Theater. Real estate developers, Poko Partners, bought the theater with the plan for a mixed use, retail and residential building; however, plans fell through and this project was then revised. The project consists of renovating 155,000 square feet into approximately 60,000 square feet of retail and 90,000 square feet of learning space for an ASCEND charter school. The new education facility will house 1,100 students from kindergarten to eighth grade. ASCEND’s plan is to add a new grade each year and are currently serving students from kindergarten to fifth grade. Their mission is to prepare economically disadvantaged, urban students for college and close the achievement gap in respect to socioeconomic backgrounds. While ASCEND was an existing organization at the time, it was relatively new at the time of the project with their first flagship 38 school opening in 2008 in Brooklyn. This project is located in the Brownsville area of Brooklyn with a poverty rate of 41.5 percent and 36.62 percent of the population over 25 years old with less than a high school diploma. While this project was scheduled to be completed in late 2011, the schedule went slightly over with the opening of the building in September 2012. In order to fund this $43.3 million project, 4 CDEs matched the cost with NMTCs. The Nonprofit Finance fund contributed $14 million; the Rose Urban Green Fund contributed $11 million, the Carver Community Development Corporation contributed: $13.3 million and ERF contributed $5 million. Goldman Sachs was heavily involved in financing the project with $12.3 million in NMTC equity and a $24.8 million senior loan. This project provides short-term impacts on the economic levels of the community by providing jobs; 160 construction jobs, 120 full-time permanent jobs, and 40 part-time jobs. This project also provides continuity of the retail corridor along Pitkin as well as provides more retail options for residents. Also, this project will provide space for working families to send their children for a better education and increase education attainment levels of the area, which is much needed per the existing levels mentioned above. Table 2 below summarizes the details of each project for a quick comparison. Table 2 Characteristics of Selected Projects Project Name Yonkers Pier Center for Urban Community Services Albany Medical Center Surgeons Pavilion McKibbin Street Industrial Center Pitkin Avenue Project CDE HEDC New Markets Inc. Local Initiatives Support Corporation Chase New Markets Corporation Empowerment Reinvestment Fund Nonprofit Finance Fund Allocation Round 2008 2007 2010 2008 2009 Year Initiated 2008 2007 2010 2007 2010 Year Completed 2009 2007 2011 2009 2012 Location Yonkers, NY New York, NY Albany, NY Brooklyn, NY Brooklyn, NY Metropolitan or non metropolitan Metropolitan Metropolitan Metropolitan Metropolitan Metropolitan Census Tract Size 4,837 3,751 3,378 661 5,520 Project Type Commercial Real Estate Real Estate Real Estate Commercial Real Estate Commercial Real Estate New or existing entity Existing Existing Existing Existing Existing Percent poverty 30% 43.9% 28.1% 8.3% 41.5% 39 of census tract Pre-project investment and economic development conditions in the target neighborhood/ property Deteriorated, limited development potential Deteriorated, limited development potential Not blighted Deteriorated, limited development potential Blighted area Nature of project Restored first- floor pavilion of the pier provides public recreation space and access to the waterfront Acquisition and renovation of a six - story commercial loft building to community facility New construction of 9 story mixed use – medical office/retail building Renovation of manufacturing building for light industrial and artisanal uses Mixed Use Real Estate Project - a charter school and a retail development Estimated Jobs Created 96 jobs created 91 construction jobs, 120 permanent jobs 135 construction jobs, 175 permanent full- time jobs, 75 permanent part- time jobs 100 construction jobs, 100 permanent full- time jobs 160 construction jobs, 120 permanent full- time jobs, 40 part-time Project completed on schedule/budget Yes Yes Yes Some budget and schedule overruns cushioned by NMTCs Slight delay Additional Public Funding Utilized HUD Section 108 loan, HUD Economic Development Initiative grant funds, Port Authority of NY & NJ funds None None Historic Rehabilitation Tax Credits, NYC Acquisition Fund Historic Rehabilitation Tax Credits, market-rate senior debt and New York City capital grants Total Project Cost $12 million $11 million $15.2 million $17.8 million $43.3 million Total NMTC Allocation $7.3 million $11 million $12.1 million $17.1 million ($4.65 million from ERF) $43.3 million * Community size, poverty rates, and MFI figures taken from 2000 Census Across projects the type of projects are all similar and consistent with the trends the GAO found in that the majority of projects were related to real estate. Another common characteristic among the projects selected were that organizations or companies involved in receiving the tax credits were existing businesses. In terms of the neighborhoods chosen the common characteristics were that all were located within metropolitan area of New York, this trend coincides with the overall trend of projects across the country where roughly 80 percent of projects are located within metropolitan areas. Economic variables measured were in line with characteristics in distressed 40 communities. CDEs benefit from focusing on projects within distressed communities because these areas will likely qualify for the Program consistently. If we compare the median case study projects’ characteristics to trends across the state of New York and the national trends; we find that the median total cost and NMTC allocation as a percentage of the total cost are slightly higher. This may be attributed to the fact that the sample consists of national CDEs. 3 of the 5 projects have both a profit and community focus. The Yonkers Recreation Pier provides space and access to the waterfront to the public along with providing retail space. Similarly, the Pitkin Avenue Project provides retail space on the lower floors for continuity of the retail corridor along the Avenue as well as provides square footage for teaching economically disadvantaged students. The Center for Urban Community Services entire purpose is to serve the disenfranchised and extend services in the immediate area where there is a high need for human services. Even though one may argue that the McKibbin Industrial Center protects local artisans and small businesses, the direct link to the targeted populations outlined by the NMTC Program is not as obvious as in the other 3 projects. Most of the project details collected lack information on the impact of the project itself on the surrounding community. One of the variables commonly used to show a project’s impact is the number of jobs created directly from the project. In this case the number of jobs created was pretty consistent and ranged from 91 to 160 construction jobs and from 100 to 170 full-time permanent positions. Only the Yonkers Pier and McKibbin Street Industrial Center specifically states that they hired local residents. If the projects hires from outside census tracts, it is likely the surrounding neighborhood economic benefits are limited depending on the size of the project. The average population of those 16 years old and over, which is an approximate of those available for work, is approximately 3,000. If we compare the number of jobs created to the community size, we find that each project on their own may not directly affect the poverty rate or employment on a large scale. In respect to financing, two of the five projects had issues early on that either required additional funding or a revised project altogether. The unforeseen structural problems in the McKibbin Street Industrial Center required GDMC to look for outside funding. By adjusting the development plan in the Pitkin Avenue project to exclude residential, the project was then eligible for NMTCs. In the last two projects, multiple CDEs allocated tax credits to each project 41 rather an individual one. Table 3 below shows how these 5 projects compare to both the state of New York and national trends from NMTC allocations from 2002 through 2010. Table 3 Case Study, New York, and National Trends Geography Percentage of Nonmetropolitan Project s Percentage of projects related to real estate Median Project Cost Median NMTCs value/ the total project cost Percentage of Multi-CDE Projects Case Studies 0% 100% $15.2 million 80% 40% New York 3% 64% $12.9 million 76% 13% National 17% 47% $5.9 million 80% 10% * Percentages calculated using analysis of CDFI Fund data Additionally, Table 3 shows how well the case studies align with the trends across the state of New York; the majority of the projects are real estate related and the median NMTCs value as a percentage of the total project cost is roughly 80 percent. The median project cost of the case studies project is slightly higher which can be attributed to the fact that real estate projects costs more on average or that project costs are higher in New York than the national average. While 64% of the projects within New York are real estate related, all of the case study projects are real estate related. This can be attributed to a couple factors. First, all CDEs associated with the case studies are national entities which may have their own trends. Second, New York has a higher percentage of projects that are real estate related. The category with the largest discrepancy is the percentage of the projects with multiple CDEs. This is likely to due to the CDEs chosen as Chase New Markets Corporation specifically seeks out to partner with other CDEs. The advantage of the case study method is that even on a small scale, we are able to observe impacts. Also, we are able to observe the additional benefit of increased access to services that may not necessarily be included in an econometric analysis. Unfortunately, there are limitations with this approach. This approach does give a snapshot of the number of jobs created and the intended benefits that a neighborhood can expect, but may not be indicative of the overall impact on the surrounding neighborhoods. These studies may be deemed successful in the eyes of the respective CDEs, but does not provide a consistent or standard measure of evaluating how these projects are affecting area in which the projects take place. The first issue in the process of choosing these projects is that by choosing successful CDEs, the geographic scopes are the same 42 across the board in that the CDEs chosen service projects across the country in all are national CDEs. While these case studies provide detailed descriptions on the potential impacts of existing and future projects, they give us limited insight to the overall impact of the program on the low- income neighborhoods across the state. Figure 3 Program Evaluation Framework * Created by Bartik & Bingham (2007) Bartik and Bingham (2007) created a framework to guide researchers in how to think about program evaluation. The figure above shows the stages of evaluation through the life of the program. Each of these six stages builds off the previous one. The first two stages focus on how the program is executed. The first stage of monitoring daily tasks takes a look at whether or not tasks are being completed efficiently and contractual obligations are being adhered to. The second stage examines whether program procedures are simple or complicated. This stage also monitors whether the implementation of the program is successful or not. The “Enumerating Outcomes” stage starts the evaluation of program outcomes. This stage explores whether program objectives are being achieved. For example, this stage would study whether the NMTC Program has met its goal of bringing investment into low-income neighborhoods. The fourth stage of the framework is similar to the third but looks at whether or not the program is working in achieving its goal. The last two stages complete more in depth analyses on costs, benefits, and the impact on the issue the program set out to address. The last stage examines the counterfactual of what would happen in absences of the program so that the outcomes associated with the intervention can be distinguished from those that would have happened regardless. While the case study method may give insight into what projects entail on a small scale, the method only enumerates outcomes rather than assesses impacts on a larger scale. The trends found within this case study must be supplemented with additional data in order to extrapolate trends for the state or country rather than be antidotal. Also, I would argue that the case study Monitoring Daily Tasks Assessing Program Activities Enumerating Outcomes Measuring Effectiveness Costs & Benefits Assessing Impact on the Problem 43 method restricts researchers in scope and keeps them stuck in the “Enumerating Outcomes” stage. The following empirical approach attempts to examine the most current data to find the impact of the NMTC Program on neighborhood variables and push the discussion of NMTC Program forward, so we can assess the impact of the NMTC Program. Chapter Three – Empirical Approach of Impact of NMTC Program Data The primary data used in the analysis comes from the U.S. Census Bureau and is broken down geographically to census tract level. The observed units are census tracts that qualified for the NMTC Program in initial allocation within the State of New York. There are 1844 individual census tracts used in this analysis. Tract level was used for two reasons. First, census tracts are the Census Bureau’s approximation for a neighborhood. They range in populations from 1,200 to 8,000, with the national average at approximately 4,000 (Census Bureau, 2013). The 2012 average for New York is close to the national average at 3,944 people, which is consistent with the national average. In order to gather data that straddles the time of the policy change, neighborhood characteristics from the 2000 Census are compared to data gathered by the 2012 5 year estimate American Community Survey (ACS), which is the most current ACS available. 2003 was the NMTC Program’s first year for tax credit allocations, so 2000 was used for the pre-policy time period and 2012 as the post-policy period. The reasoning behind a long post policy period was that the majority of QLICIs fund real estate projects, which are long term in nature. The 2010 Decennial Census gives us the number of people who live in the United States rather than an idea of how people live, which the ACS does. The American Community Survey is a statistical survey that samples a small percentage of the population on an annual basis in order to provide communities, researchers, and local and federal governments with information to best plan their investments and services. The Census Bureau selects roughly 3. 5 million addresses randomly each year. Each selected household is then asked to fill out either a paper survey or an online survey. An example is included in the Appendix. Households are required by law to complete the survey. The Bureau then follows up with those that do not send in a completed survey by either a visit to the address or by phone. This survey method minimizes the possibility of self-selection bias and 44 nonresponse bias that plague many data sets collected through surveys. The ACS asks sampled households about: age, sex, race, family and relationships, income and benefits, health insurance, education, veteran status, disabilities, where you work and how you get there, and where you live and how much you pay for some essentials. The Census Bureau then uses the survey data to create 1, 3, and 5 year estimates. For this study 5 year estimates were used. The reason for this is that the Census Bureau only reports census tract information on a 5 year level in order to protect the privacy of those surveyed. This set of data uses 60 months of collected data, provides data for all available areas, and is most reliable. Unfortunately, a disadvantage to this set is that because it uses 60 months of collected data, it is also the least current of the three. While the 5 year estimates use the oldest data among the 3 estimates, they have the largest sample size and are the most reliable. The first ACS available on the U.S. Census Bureau website was collected after the NMTC Program was enacted. For this reason, the 2000 Census was used to set the baseline for neighborhood characteristics. Unfortunately the data set needed some adjustments before it could be analyzed. In 2010 The U.S. Census Bureau updated census tract borders. They split up tracts with populations over 8,000 individuals, joined tracts with populations less than 1,200 with neighboring tracts, and made small boundary adjustments. In order to process data and overcome this obstacle the 2010 Census Tract Relationship file was used in conjunction with the Bureau’s record layout information for the census tract relationship files to adjust tracts. For tracts that split, the 2000 tract data was used as the baseline like the majority of tracts. For tracts that were merged, a weighted average of the neighborhood characteristics was created based on population size. For example, the 2010 census tract 36103201100 is a combination of 2000 census tracts 36103145500 and 36103145601. Since both 2000 census tracts were absorbed into tract 36103201100 completely, the ratios of their respective populations are used to weight neighborhood characteristics. First, the 2000 baseline poverty rate is created using the 2000 poverty rate for 36103145500 is multiplied by its population of 184 over 5209, which is the sum of population counts for both census tracts. The same is done for tract 3610314560. Then, these the results are summed for tract 36103201100’s baseline poverty rate. 445 of 1,931 qualifying tracts were affected by this issue. The Bureau calculated the percentage of the 2000 population contained within the 2010 population for the new tract, which was then multiplied by the 45 neighborhood characteristic of that tract to calculate the baseline. Part of some 2000 tracts were split with a portion of the original tract being merged with other tracts. Table 3 Names, Descriptions, and Sources for Variables Name Description Source Poverty Rate T percentage of individuals under the poverty threshold within each census tract for the 2010 5 year estimate and the 2000 Census U.S. Census Bureau (2000 Census & American Community Survey 5 year estimates 2006-2010) Treatment Dummy variable equal to 1 if the tract received investments due to NMTCs CDFI & Novogradac Files for NMTC Transactions in New York Unemployment Rate The percentage of the population 16 years old and over classified as unemployed U.S. Census Bureau (2000 Census & American Community Survey 5 year estimates 2006-2010) Migration The percentage of the population 1 year old and over that have moved into the census tract from outside the county U.S. Census Bureau (2000 Census & American Community Survey 5 year estimates 2006-2010) Note: The U.S. Census Bureau classifies those as unemployed to be those 16 years old and over that are available for work, currently looking for employment, and without a job The Census Bureau uses national income thresholds based on family composition and size to determine which individuals are in poverty. For individuals who live with family members, the poverty status is based on total family income. As a result, all family members are assigned the same poverty status. For individuals that do not live with any family member, their income is compared to the respective category. Income is defined as “money income,” which includes earnings, unemployment compensation, workers' compensation, Social Security, public assistance, veterans' payments, pension or retirement income, interest, dividends, rents, royalties, income from estates, trusts, educational assistance, alimony, child support, assistance from outside the household, and other miscellaneous sources before taxes. Income in this case does not include food stamps, housing subsidies, and capital gains or losses. This is one of disadvantages to using this data set as many of the individuals in qualifying census tracts receive a form of assistance outside what is considered to be “money income.” Each individual or family is designated one out of 48 different thresholds. A list of all poverty thresholds for 2012 is provided within the Appendix. These thresholds are adjusted annually for inflation using the Consumer Price Index for All Urban Consumers. Because of this no additional calculation was done to adjust for inflation within this paper. The total family income is compared to these thresholds. If a family or individual’s income is greater than their respective threshold, then they 46 are not in poverty and vice versa. The outcome or dependent variable in this analysis is the percentage point change in poverty status from 2000 to ACS 2012 5 year estimate. Methodology The methodology used to analyze the question of the impact of the NMTC Program on New York is based on the previous work done by David Card and Alan B. Krueger in Minimum Wages and Employment: A Case Study of the Fast Food Industry in New Jersey and Pennsylvania (1994). While the previous study analyzed minimum wage changes in New Jersey using data collected pre and post wage change, this study focuses on the change in poverty rate using data collected pre and post disbursement of NMTCs. This study also takes into account work completed by Anderson and Wasserman (2000) in respect to the poverty rate. They use numerous variables to define the function of poverty, which includes dummy variables to denote the city in which the observation lies. Poverty rate was chosen as the dependent variable in this study due to the nature of the economic indicators politicians use. The poverty rate is also one of the indicators used to qualify for the program. Due to regional differences between employment and basic economy sectors, four dummy variables have been created to identify census tracts located in the four regions; Northwest, Northeast, Southwest, and Southeast. In order to distinguish tracts that have received investment as a result of tax credit allocations, a treatment variable has been added as well. The treatment group is defined as census tracts that have received investments due to NMTCs from 2003-2010, while the control group has not. Freedman’s study is one of the few that has specifically analyzed the NMTC Program using econometric models. He used random assignment to create the treatment and control groups, which is questionable as the GAO had previously reported the allocation process is not random (GAO, 2009). The CDFI Fund records where tax credit allocations have been used to gain investment. This same data set was used to choose the projects examined in Chapter 2’s cases study analysis. We are able to track which census tracts have received investment due to the NMTC Program. We can then use this data to create the treatment and control groups for this analysis. Like many policies analysis, this paper takes the approach of examining the change in neighborhood characteristics pre and post policy enactment to shed light on whether the change can be 47 attributed to the policy. In order to compare social characteristics, this paper adopts a difference in difference model to explain the change of the poverty rate between 2000 and 2012 within census tracts that received investment due to tax credits. Table 4 provides means of the significant variables affecting poverty included in Anderson and Wasserman’s analysis separated for the two groups. 48 Table 4 Means of Key Variables Low Income Communities Variable Tracts Receiving Investment Tracts Not Receiving Investment t statistic 1. Means in 2000 a. Poverty rate 32.84 26.39 -5.27 b. Unemployment rate 6.97 6.55 -0.90 c. Share of population with less than a high school degree a 37.36 33.03 -3.38 d. Share of population with a Bachelor's degree or higher a 15.18 15.16 -0.02 e. Share of population that identifies as African American 37.38 26.92 -3.68 f. Share of population under 18 years old 26.97 27.32 0.47 g. Share of population 65 years old and over 10.83 11.34 0.87 h. Share of population that have immigrated to tract from previous year b .226 .223 -2.320 2. Means in 2008-2012 a. Poverty rate 28.23 25.53 -2.09 b. Unemployment rate 7.00 7.12 0.34 c. Share of population with less than a high school degree a 24.60 23.78 -0.69 d. Share of population with a Bachelor's degree or higher a 25.63 20.93 -3.50 e. Share of population that identifies as African American 32.19 25.68 -2.42 f. Share of population under 18 years old 22.40 23.95 2.02 g. Share of population 65 years old and over 12.08 11.48 -1.03 h. Share of population that have immigrated to tract from previous year b 6.445 5.080 -2.402 3. Differences a. Change in poverty rate -4.61 -0.86 3.68 b. Change in unemployment rate 0.57 0.02 1.04 c. Change in migration 6.178 4.857 -2.370 Note: a Population for these variables is defined as those 25 years old and older rather than the total population estimate of the census tract used for the other variables b Population for these variables is defined as those 1 year old and older rather than the total population estimate of the census tract that have moved within the previous year from outside the county The last column of Table 4 provides the t statistic for a test of equality of means between the control and treatment groups. Using this table we find that in 2000, the treatment group has statistically significant higher averages of the poverty rate, the share of population that have less 49 than a high school degree, share of population that identifies as African American and the share that has immigrated to the tract from the previous year. Comparing these to the averages of the 2008-2012 ACS estimates, the changes between groups are the treatment group’s average in the share of the population with less than a high school degree is no longer statistically different and the share of those with the a bachelor’s degree or higher is now statistically significant. Comparing the averages between the groups and time may indicate that that there are changes in the composition of populations within New York census tracts. Figure 5 – Distribution of Poverty Rates 0.00% 5.00% 10.00% 15.00% 20.00% 25.00% 30.00% 35.00% 40.00% Distribution Poverty Rates in 2000 Distribution of Poverty Rate in 2000 Control Group Treatment Group 50 Figure 5 above shows the distribution of poverty rates for the two groups. In 2000, we see that the treatment group had a greater percentage of census tracts with poverty rates over 40 percent. In the second graph, we see the treatment group’s poverty rates drop and shift left, while the distribution of the control group’s poverty rate remain similar to those in 2000. We then use this data is estimate the difference-in-difference model with the hypothesis that tracts within the treatment group will have a statistically negative relationship with the change poverty rate. The assumptions of the model are similar to that of the Ordinary Least Squares (OLS) model. Studenmund (2006) states the first assumption is that the regression model has a linear functional form, correctly specified, and has an additive error term. The second assumption is that the error term has a population mean of zero. The third OLS assumption is that the explanatory variables are not correlated with the error term. The fourth assumption states observations of the error term are not correlated with each other. The fifth is the variance of the error terms is constant. The sixth is that perfect multicollinearity exists between explanatory variables. The last OLS assumption is that the error term is distributed normally. In addition, the DID model requires a parallel trend assumption, which assumes that the outcomes in the treatment and control groups would follow the same trends in the absence of the treatment. This study also assumes that the 0% 5% 10% 15% 20% 25% 30% 35% Distribution Poverty Rate Percentages in 2008-2012 Distribution of Poverty Rates in 2008-2012 Control Group Treatment Group 51 composition of the both the treatment and control group are unchanged or have minimal change between periods. Using the repeated cross-section data the ACS provides, we would need to know the future treatment status of the census tracts observed in the initial period, period t= 0. Knowing the future treatment status of each census tract in the initial period may be difficult to ascertain due to changes in income. Since we are sampling a changing population at different points in time, the distribution of the neighborhood variables are likely to change over time. Thus, the identical distribution assumption is no longer valid, so we then assume observations are independent, not identically distributed. In this analysis we have two time periods: period 0, which is the period before the enactment of the NMTC Program, and period 1, which the period after the Program was enacted. 2000 would be the period before the Program and the 2012 5 year estimate would be the period after the Program. We also have individual census tracts affected by the policy intervention, d = 1 and census tracts that were not affected by the policy intervention, d = 0. Because NMTCs are allocated credits per census tract, the following analysis is on the census tract level, where individuals are the different census tracts within New York that qualify for the NMTCs. For this analysis, Y is the change in the poverty rate and is the percentage of the population with poverty status in the last 12 months. Y dt = individual change in poverty rate in period t if treatment status is d. We observe the Y 0 , Y 1 , D with D the observed treatment status in period 1. In period 0, d can only take the value of 0 as the policy intervention has yet to occur so that Y 0 = Y 00 . In period 1, d = 0, 1 so that Y 1 = DY 11 + (1 – D) Y 01 . The causal effect of the intervention on an individual census tract is equal to the individual treatment effect defined as Y 11 – Y 01 . Unfortunately since we can observe either Y 11 or Y 01 and not both, we cannot observe the individual treatment effect. Because Y 11 and Y 01 are the potential outcomes and one of the potential outcomes is a counterfactual, we can use the Average Treatment Effect (ATE) to measure the average causal difference in outcomes under the treatment and under the control. The ATE method uses random assignment, like that seen in Freedman’s analysis, to apply the intervention in period 1. The difference of the average outcome of the treated and control group results in then the ATE. Unfortunately, the NMTC allocations are not chosen at random as the GAO reports. We can see 52 this is not the case by comparing the average poverty rate before the NMTC Program. The averages of the poverty rate percentages are listed in Table 5 below. The differences within the table are the simple percent point differences between the groups and periods. Table 5 Comparison of the Poverty Rate between Groups Period Before Period After Period Differences Control 26.38979 25.52674 0.86305 Treated 32.83519 28.23017 4.60502 Group Differences -6.445403 -2.703436 -3.74197 We are most interested in comparing the average poverty rates before and after the implementation of the NMTC Program. This is equivalent to E (Y 0 │D = 1) – E (Y 0 │D = 0) = E (Y 01 │D = 1) – E (Y 00 │D = 0). Table 4 shows that these averages are 32.836 and 26.389, which implies there may be a lack of random assignment. We then assume that the selection for the treatment can depend on the initial status of each census tract in 2000 but is dependent of the change in the non-treated outcome, so that Y 01 - Y 00 ⊥ D. This means that the change in the average poverty rate would follow the same trend for all low-income census tracts if the NMTC Program had not been administered. With this new assumption, we have E(Y 1 - Y 0 │D = 1) - E(Y 1 - Y 0 │D = 0) = E(Y 01 - Y 00 │D = 1) - E(Y 11 - Y 01 │D = 1) - E(Y 01 - Y 00 │D = 0) = E(Y 11 - Y 01 │D = 1) If we rewrite the left hand side we have E(Y 1 │D = 1) - E(Y 1 │D = 0) - E(Y 0 │D = 0) = E(Y 11 - Y 01 │D = 1) The left hand side of the equation above is the difference in average outcomes between the treated and controls in period 1 minus the difference in the average outcomes between treated and controls in the initial period, period 0. By subtracting the average outcome between the two groups in period 0, we are correcting the possibility of selective application for the treatment. The left hand side of the equation equals the difference – in –differences. The right hand side of 53 the equation is then the average treatment effect on the treated individuals so that the basic linear model is: Y = α + γ 1 T + γ 2 D + βT * D + ε T is a dummy variable denoting the time period. The time variable appears in this equation for heteroskedasticity to figure out whether the unconditional error variance has changed with time. (Wooldridge, 2012). D is the dummy variable for the treatment. For this study we are interested in the change in poverty rate that is correlated with tracts that have received NMTCs, so we define the Y as the change in the poverty rate between periods. Because we are analyzing the change, we drop the time variable so that the OLS regression model becomes: ΔY = α + βD + ε (1) We then consider two conditional diff-in-diff estimators that make different assumptions on the selection for the treatment. The first assumption is that without the QLICIs, the poverty rate in the State of New York would have been the same in all qualifying low census tracts both within all counties. A dummy variable is added to the basic equation to control for location differences between regions. Significant regional differences include difference within the type of primary industry within the region and regional employment market differences. Under this assumption we use the following regression, where β remains the same estimate for the correlation coefficient of the treatment dummy and δ 1 to δ 3 is the correlation coefficient estimator for these regional dummy variables, where 1 denotes a tract within the respective region. Tracts are assigned one of four regions; northwest, southwest, northeast, and southeast, based on the county in which they are located. A map of the New York counties and their respective region has been included within the Appendix. We then add the regional dummies to create Model 2. Δ Poverty Rate = α + βD + δ 1 NW+δ 2 SW+ δ 3 NE + ε (2) The next assumption is that without investment due to NMTCs, the poverty rate in New York would have been the same for all qualifying census tracts of the same level in period 0. We then run the change in poverty rate on the treatment dummy and poverty in period 0. This baseline poverty rate controls for initial level of poverty. 54 Δ Poverty Rate = α + βD + δ 1 Poverty Rate 0i + ε (3) Lastly, model 4 includes both conditional assumptions. Since the poverty rates in 2000 are not highly correlated with the regional dummies, models 2 and 3 are combined to create model 4. Δ Poverty Rate = α + βD + δ 1 NW+δ 2 SW+ δ 3 NE + δ 4 Poverty Rate 0i + ε (4) Results A summary of all work completed has been included in the Appendix section, which includes the STATA output for all models. Table 6 shows the models for change in poverty rate. As previously stated, the hypothesis of this study is that we would see a statistically significant negative correlation between the change in poverty rate and the treatment dummy. We can use Table 5 to populate the first model for the change in poverty, where Δ Poverty Rate = α + βD + ε, as the basic model is directly comparable to the simple differences in differences. Table 6 - Reduced Form Models for Change in Poverty Rate Model Independent Variable (1) (2) (3) (4) Treatment Dummy -3.741 -3.115 -1.912 -1.634 (1.016) (.979) (.963) (.936) Control for regions a no yes no yes Control for initial baseline no no yes yes Notes: Standard errors are given in parentheses. The sample consists of 1844 census tracts with available data on poverty rates in both time periods. The dependent variable in all models is the change in poverty rates. The mean and standard deviation of the change in poverty rates are -1.098 and 10.636 respectively. All models include an unrestricted constant reported in within the STATA documentation in the Appendix. a Dummy variables were created to denote census tracts located within various regions of New York Table 6 does not make any allowances for other variations in poverty change. Model (1) shows a statistically treatment shows the most basic regression model, where we regress the change in poverty rate on the treatment dummy. Model 1 shows that the average difference in the poverty rate change between tracts that receive QEIs and those that do not is -3.41 percent with a 55 standard error of 1.016. Model 2 shows a similar result with the addition of a dummy variable for the metropolitan tracts. The treatment dummy coefficient only drops slightly to -3.661 with a standard error of 1.014. We added the assumption that without investment due to NMTCs, the poverty rate in New York would have been the same for all qualifying census tracts of the same level in period 0 in Model 3. This model shows a significant drop in the magnitude of the treatment coefficient and the effect of the treatment is not as big as we had seen in the first two models. While significant the coefficient estimate of the poverty rate baseline is -.257 with a standard error of .018. With the additional control variables, the t statistics for all models are high enough to reject the null hypothesis, the treatment effect estimator equivalent to or less than 0, at the 95% confidence level. In order to put these results into context I reference an ACS brief studying the poverty rate for the same time period as this study. The study focused on national and states poverty rate trends comparing 2000 and 2012. Impact of the NMTC Program on Other Variables In order to gain more insight to how the NMTC Program has impacted low-income neighborhoods we use the same models as we did for the change in poverty rate except we change the dependent variables. First, the change in unemployment rate was used as the dependent variable. The expectations were that the same trends for the change in poverty rate would be seen for the change in unemployment. This assumes that the reason we were seeing a decrease in the poverty rate were that the NMTC funded projects were creating jobs as the projects in the case study suggested. The poverty rate thresholds are adjusted for inflation on an annual basis, so we are looking for reasons on how the Program impacted the annual income of affected tracts. 56 Table 7 - Reduced Form Models for Change in Unemployment Rate Model Independent Variable (1) (2) (3) (4) Treatment Dummy -.544 -.459 -.185 -.219 (.525) (.525) (.343) (.342) Control for regions a no yes no yes Control for initial baseline no no yes yes Notes: Standard errors are given in parentheses. The sample consists of 1844 census tracts with available data on poverty rates in both time periods. The dependent variable in all models is the change in poverty rates. The mean and standard deviation of the change in poverty rates are -1.098 and 10.636 respectively. All models include an unrestricted constant reported in within the STATA documentation in the Appendix. a Dummy variables were created to denote census tracts located within various regions of New York Unexpectedly we see that the magnitude of the treatment affect is small if any. The t statistics for all models are low, so that we cannot reject the null hypothesis of a zero unemployment effect of the NMTC Program. For the next model, Table 4 was examined to find other variables that are affected by the NMTC Program. Looking at the second period, we see the most drastic differences between groups outside of the poverty rate to be the share of the population that identifies themselves as African American and those with a bachelor’s degree. Table 8 - Reduced Form Models for Change in Migration Model Independent Variable (1) (2) (3) (4) Treatment Dummy 1.321 1.440 .600 .616 (.557) (.556) (.463) (.464) Control for regions a no yes no yes Control for initial baseline no no yes yes Notes: Standard errors are given in parentheses. The sample consists of 1844 census tracts with available data on poverty rates in both time periods. The dependent variable in all models is the change in poverty rates. The mean and standard deviation of the change in poverty rates are -1.098 and 10.636 respectively. All models include an unrestricted constant reported in within the STATA documentation in the Appendix. a Dummy variables were created to denote census tracts located within various regions of New York Freedman’s analysis concluded that the decrease in poverty and unemployment rates associated with the Program may be due to changes in population composition. Roger M. Groves provides a reason to why populations may be changing and how that would affect the poverty rate. He states 57 that loopholes within the regulations of the NMTC Program has created “subsidized gentrification” because the Program has allowed the development of projects that focus on a wealthier population that can potentially move into the area rather than the existing low-income residents. The hypothesis then is that the change in migration and the treatment dummy would be significant and positive. In order to investigate this, the dependent variable in the previous models is adjusted to the difference in those living in a different house in the previous year. The results find a statistically significant correlation between the treatment dummy and the change in migration when we do not control for the initial migration baseline. We cannot reject the null hypothesis that the mean of the treatment coefficient estimate of the treatment dummy is statistically significant different from zero. Comparison of Standard Errors In order to examine whether the standard errors reported above are understated, we employ a clustering method. In 2004, Bertrand, Duflo, and Mullainathan published “How Much Should We Trust Difference in Differences.” They found that conventional differences in differences, like the one preformed above, understate the actual sampling variation of the coefficient estimators. The authors also found that difference in difference studies that were more successful calculating standard errors accounted for clustering (Bertrand, Duflo, Mullianathan, 2004). Let us look at model (2) for comparison purposes as this model took into consideration difference between regions and therefore cluster the data into 8; 4 four different regions times for both control and treatment groups. We now include all regions for the dummy variables and suppress the constant for Model 5. We first use the correlation factor method, which adjusts the OLS standard errors for correlations between errors within clusters of observations. The variance of errors for the OLS model does not distinguish between idiosyncratic errors and cluster specific error. To estimate the variance due to idiosyncratic error, we take the sample variance of the OLS residuals from Model 5. We then subtract the cluster averages from the residuals and divide by n-G-1, were n is the total number of individuals and G is the number of groups. This will give us the estimate of the idiosyncratic error. We then estimate the fraction of the variation due to cluster specific error so that 𝜌 � = 𝜎 � 2 − 𝜎 � 2 𝜎 � 2 , where 𝜎 � 2 is the sample variance of the OLS residuals and 𝜎 � 2 is the idiosyncratic error. To calculate 𝜎 � 2 cluster averages of the OLS residuals were subtracted from the existing OLS residuals. The squares of the adjusted residuals were then 58 summed and divided by n, where n is the total number of census tracts. The simplest method was used to calculate the variance of the adjusted residuals since the cluster effect is small. Using the OLS residuals from Model 5, we find 𝜎 � 2 =103.7518, 𝜎 � 2 = 103.6848, and 𝜌 � =.000646. With an average cluster size of 230.5, the correction factor is 1.1481, so that the cluster-corrected errors are 1.0715 times larger for a region-level regressor than those computed by the standard OLS technique in Model 5. An alternative to clustering is the bootstrap method, which uses a nonparametric bootstrap to estimate the sampling distribution in the model. This process treats the observed data set as a complete population and draws a new, simulated sample from it, picking each observation with equal probability with replacement and then re-running the estimation. This method works due to the law of large numbers. The larger the number of sample size n are, the greater the accuracy of the relative distribution of the estimates. As the number of repetition of sampling reaches infinity, the expected value of the estimated sample variance of the sample mean converges to the expected value of the true population. To calculate the nonparametric bootstrap of the standard errors the OLS Model 2, we use our existing cross sectional y i, xʹ i , I = 1,…., n of observations on the change of poverty and 5 independent variables. xʹ i = (1 xʹ 1i ) so that the OLS estimator is 𝛽 ̂ = � � x i n i= 1 x i ′ � − 1 � � x i n i= 1 y i � We then draw a sample of y * nb, x * ʹ nb data from the empirical distribution y 1, xʹ 11 , …, y n, x * ʹ 1n with replacement. Then, we approximate the OLS estimator using the new samples we pulled. We repeat the process 5,000 times with a sampling size of 500 and compute the estimated variance of the various correlation coefficients. The revised standard errors using the correction factor and bootstrap method are reported in Table 9. 59 Table 9 – Comparison of Standard Errors using Alternative Methods Variable Coefficient OLS S.E. Correction Factor S.E. Bootstrap S.E. Treatment -3.1152 .9795 1.0495 2.2826 Northwest 4.2447 .8259 0.8849 1.9764 Southwest 3.5748 .6064 0.6497 1.7268 Northeast 3.5377 1.1263 1.2068 3.4867 Southeast -2.7312 .2886 0.3092 1.6026 From Table 9, that the standard OLS model did underestimate the standard errors. Even with the adjusted standard errors from correction factor method, we continue to have a statistically significant negative correlation between the treatment dummy variable and the change in poverty rate for a one-sided test at a 95 percent significance level. However, the bootstrap method provides significantly larger standard errors that make the treatment variable statistically significant at the 90 percent confidence level rather than 95 percent. Chapter Four - Conclusion & Policy Recommendations The case study completed in Chapter 2 showed that the New Market Tax Credit Program has made an impact on surrounding community by creating both construction and permanent jobs. These projects have also served an additional purpose by creating greater access to health, human, and educational services in underserved communities. The results from the econometric analysis show that there is a negative and statistically significant correlation between receiving investment as a result of the NMTC Program and the poverty rate. In the first two models the being in the treatment group is correlated with over a 3.5 percent drop in poverty, keeping all other independent variables constant. Even when we control for the baseline poverty rate, the treatment dummy is still statistically significant. While the magnitude of the correlation drops from -3.661 percent to -1.910 percent from model 2 to model 4, which controls for the baseline poverty rate, we still see a statistically significant and negative correlation. These findings support the extension of the New Market Tax Credit Program. These findings are consistent with Freedman’s study, but show a larger magnitude of the treatment effect on poverty rate. Even with the adjusted standard errors from the correction factor and bootstrap method, the treatment variable remains statistically significant at varying confidence levels. This supports the findings 60 from the OLS model that the census tracts receiving investment due to the NMTC Program see a drop in the poverty rate. There are also limitations to this empirical model, the biggest one being that the difference-in- difference model makes strong assumptions. One of the key assumptions of this empirical method is the difference between the treatment and control groups would have remained constant in the absence of the treatment. Unfortunately, we are unable to test this. A related concept is that of Ashenfelter's dip, where individuals encounter an idiosyncratic shock just prior to the treatment. Unfortunately, there is a limitation within the data available to test for such as dip. The first year the ACS provided 5 year estimates is in 2008, which is after the first round of NMTC s allocations in 2003. The only other poverty rate data available on the census tract level is provided through the decennial census. Poverty rate data collected every 10 years is unlikely to capture a shock right before the intervention whereas data collected annually could have. This analysis has also shown that the treatment dummy does not have a statistically significant relationship with changes in the employment rate. Further research needs to be completed on why the poverty rates are decreasing but the unemployment is not at a statistically significant level. The results may infer wages and/or quality of jobs of residents within the treatment census groups are increasing at rate higher than inflation. Another theory is that those hired by NMTC funded projects are spending at local shops that are increasing their revenue without hiring additional workers. For example with the Albany Medical Center Surgeon’s Pavilion, the surgeons may live outside low income census tracts, but may contribute to the local economy by buying lunch in the area. This would increase the income of existing businesses without necessarily affecting the unemployment rate. Another reason this is occurring is that the diff-in- diff models used suffer from a specification bias. Further analysis also needs to be done to truly distinguish the effects of the program from other factors that may have influence on the poverty rate including other economic development initiatives. There has been much criticism with the difference-in-difference model in respect to whether the results should be trusted. In addition, qualitative impacts such as social change occur and are rarely accounted for in qualitative studies. There is a need to record and demonstrate how 61 these qualitative impacts affect communities to provide evidence for continuing to fund the NMTC Program and other economic development initiatives. This analysis adds to the support for the continuation and shows the negative impact on poverty that the Program has in low- income neighborhoods. However, a more comprehensive evaluation method, whether a synthesis of qualitative and quantitative analyses or simply an improved econometric method, would be needed to convince the harshest critics of the Program. Policy Recommendations Both the qualitative and quantitative analyzes show that the NMTC Program has had on a positive impact the surrounding census tracts, so the first policy recommendation is to continue the NMTC Program with additional monitoring. The case studies provide antidotal information on the existing and potential impacts of NMTC funded projects within New York. The information gathered through the case study analysis supports the continuation based on the jobs created and retained and the additional access to services in underserved communities. The reason increased monitoring is needed to ensure we are tracking the creation of jobs rather than relocation of jobs. The relocation of jobs may make it easier for certain populations to get to work, but the overall impact on poverty within New York will be unaffected as there is zero impact on the total net number of jobs. Using the example of the Center for Urban Community Services, the number of jobs created rather than relocated is unclear as some of those jobs were relocated from the existing business. However, a significant impact on the neighborhood is the relocation of services that brought human services closer to populations in high demand of these services. Because these populations tend to be disadvantaged financially, the closer the services, the less transportation burdened they are. Tracking daily activities, such as the change in the number of clients serviced, would supplement the limited picture that job creation numbers alone provide. The empirical analysis provides evidence to posit the continuation of the NMTC Program; however, it is not as strong as the observations within the case study. The data and results alone may not give a convincing argument for skeptics looking for the most financial return or impact for public spending. Both the case study and econometric model combined should provide the convincing argument that the Program has positive impacts on neighborhood characteristics and 62 the neighborhood economy. The first recommendation is to extend the program at the current budgeted amount. The reason for specification is that the investors are currently paying seventy to eighty cents per dollar of tax credit. The history of the NMTC market does not support raising funding levels. While this paper did provide evidence that the Program has a statistically significant negative correlation with the change in the poverty rate and supports an extension of the program, further analysis is needed to distinguish whether the impacts of Program are consistent across different populations in poverty. For example, would the NMTC Program have a similar impact on a census tract with a poverty rate of 30 percent and with a poverty rate of 100 percent? Also, as the Program matures, further analysis should be completed to examine whether the negative correlation found in this study was only a short term effect or not. The second policy recommendation is for economic development programs to include accountability measures that loop back to the criteria used to disseminate the program in the first place. If there is reason and evidence to use the eligibility criteria for applicable federal funding to target a specific issue, we should evaluate how the resulting affects within the context of the initial qualifying criteria. For example, in the NMTC Program eligible census tracts are chosen based on poverty rates and median family income. Low-income tracts were targeted because they tend to be at a disadvantage when seeking investment and the hope of the program is to change that status. If this is the case, then the evaluation of the impact of the Program should not stop at whether or not the investment has resulted from the Program, but whether the Program has addressed the issues that categorized these tracts as disadvantaged and qualify for the Program is the first place. The next step of program evaluation according to Bartik and Bingham’s framework answers the question of whether it has made an impact on the poverty levels or MFI that distinguished these census tracts from others in the first place. The third policy recommendation is to expand data collection. The most recent CDFI survey of CDEs was the first year in which job creation was included. Expanding data collection will allow policy makers and researchers to ascertain the impact of the NMTC Program on various aspects of poverty easier. While this study found positive impacts by creating construction jobs, data on 63 the number of permanent jobs created versus moved was ambiguous. For example, the Center of Urban Community Services, the Albany Medical Center Surgeons Pavilion, and the Pitkin Avenue Project all involved moving an existing workforce to a new space. The permanent job creation figures are likely to be overstated, so expanding data collection to decipher which jobs were created versus relocated would provide more accurate figures. Another area where more data is needed is additional benefits of the project outside of job creation. Collecting benefits would give a complete picture of what the Program can offer neighborhoods that are not captured by the income related variables. More data may give researchers and policy makers more specific ideas of how the Program is impacting neighborhoods. The previous recommendations lead to a broader recommendation, which is to create rigorous evaluation methods and set clear benchmarks at the initial design phase for economic development programs, especially federal programs, in order to effectively evaluate programs for the greatest impact. Evaluations are needed at the federal for one main reason, scale. State and local economic development spending do not compare to the amount the federal government’s for economic development programs, which was over $15 billion (Bartik, 1994). The magnitude of federal programs allow for them to have a significant impact on state or national economies, where state and local programs may not. The question of cost of funding evaluations can be spread throughout the country rather than overwhelm a smaller entity. Bartik suggests that funding for evaluations need to come from multinational consortiums of economic development agencies or national organizations, such as the National Association of State Development Agencies or the National Council for Urban Economic Development. I would argue that large nonprofit organizations be added to that list. Nonprofit organizations, such as the Bill and Melinda Gates Foundation, have the funds available to assist in evaluating these programs. We are seeing a call to partner between sectors to solve urban issues; there is no reason we cannot apply this method to funding evaluations of economic development to alleviate urban poverty. While there are limitations to this model, this paper hopes to continue and posit Batik’s argument that “that better evaluation is essential if economic development programs are to play a major role in the American economy.” 64 BIBLIOGRAPHY Abravel, Martin D., Pinduse, Nancy M., and Brett Theodos. Evaluating Community and Economic Development Programs: A Literature Review to Inform Evaluation of the New Market Tax Credit Program. Washington, DC: Urban Institute, 2010. Abravel, Martin D., Pinduse, Nancy M., and Brett Theodos. Analysis of Selected New Market Tax Credit Projects. Washington, DC: Urban Institute, 2007. Anderson, John E. and Robert W. Wassmer. Bidding for Business: The Efficacy of Local Economic Development Incentives in a Metropolitan Area. Kalamazoo: W.E. Upjohn Institute for Employment Research, 2000. Bartik, Timothy. “Better Evaluation Is Needed for Economic Development Programs to Thrive.” Economic Development Quarterly (1994): 99-106. Bernanke, Ben. “The Community Reinvestment Act: Its Evolution and New Challenges.” Speech, from Community Affairs Research Conference, Washington, D.C., March 2007. Bertrand, M., Duflo, E. and Sendhil Mullainathan. "How Much Should We Trust Differences-in- Differences Estimates?," The Quarterly Journal of Economics, MIT Press, vol. 119, No. 1(February 2004):249-275. Bishaw, Alemayehu. Poverty: 2000 to 2012. (Washington DC: U.S. Census Bureau, 2013), 1-16. Blakely, Edward J. and Nancey G. Leigh. Planning Local Economic Development: Theory and Practice. 4 th ed. Los Angeles: SAGE Publications Inc., 2010. Card, David and Alan B. Krueger. “Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania.” American Economic Association, Vol. 84, No. 4 (Sep., 1994): 772-793. Center for Urban Community Services. About Us. http://www.cucs.org/about-us Central Intelligence Agency. The World Factbook. https://www.cia.gov/library/publications/the- world-factbook/geos/us.html. Community Development Financial Institution. NMTC Program Allocatees States Served. http://cdfifund.gov/impact_we_make/nmtc_state_reports.asp Cuomo, Andrew. New Markets: The Untapped Retail Buying Power in American’s Inner Cities. Washington, DC: Government Printing Office, 1999. Daniels, B. and Economic Innovation International, Inc. Market Assessment and Investment Strategy for a Maryland Community Development Fund. Boston: 2005. Federal Reserve Board. Community Reinvestment Act (CRA). Last modified August 2013. http://www.federalreserve.gov/communitydev/cra_about.htm 65 Florida State University. Economic Base Theory. Date accessed December 1, 2013. http://mailer.fsu.edu/~tchapin/garnet-tchapin/urp5261/topics/econbase.htm Freedman, Matthew. “Teaching new markets old tricks: The effects of subsidized investment on low-income neighborhoods.”Journal of Public Economics 96 (2012): 1000-1014. Gaither, Marisa. “New Opportunities for Equity: A Critical Examination of Community Development Venture Capital.” MA thesis., Massachusetts Institute of Technology, 2002. Geographic Products Branch United States Census Bureau. “Census Tracts.” https://www.census.gov/geo/education/pdfs/CensusTracts.pdf Gorr, Wilpen L. and Kristen S. Kurland. GISTutorial 1: Basic Workbook. Redlands: ESRI Press, 2008 Greenpoint Manufacturing & Design Center. “Buildings that Work: 221 McKibbin Industrial Center.”Powerpoint presented at American Planning Association Conference, New Orleans, LA, April 2010. Groves, Roger M. "The DeGentrification of New Markets Tax Credits" Florida Tax Review 8.2 (2008): 214-265. JPMorgan Chase & Corporation. “New Markets Tax Credit (NMTC) Investments.” Date accessed December 6, 2013. https://www.chase.com/commercial-bank/finance/new- markets-tax-credit. Lu, Yang. “Lecture 4: Multicollinearity, Serial Correlation, and Heteroskedasticity.” Lecture, PPD 558 from the University of Southern California, Los Angeles, February 11, 2013. “NDC Takes First Place for Yonkers Pier in Novogradac Community Development Awards.” National Development Council. Date accessed December 6, 2013. http://nationaldevelopmentcouncil.org/blog/?p=965. New Markets Support Company. “LISC's New Markets Tax Credit Program FAQ.” Date accessed December 6, 2013. http://www.newmarkets.org/section/lisc_newmarkets/faq#how_cdes New Markets Tax Credit Coalition. “New Markets Tax Credit.” Date accessed December 6, 2013. http://waysandmeans.house.gov/uploadedfiles/new_markets_tax_credit_coalition.pdf New York City Economic Development Corporation. “Unlocking New Markets.” PowerPoint presented at Annual New Markets Tax Credits Conference of the New York City Economic Development Corporation, New York, NY, January 2012. Nonprofit Finance Fund. “What we do.” Date accessed December 6, 2013. http://nonprofitfinancefund.org/about-nff/what-we-do . Novogradac & Company LLP. “NMTC Investments in the State of New York.” Last modified May 2013. 66 http://www.novoco.com/new_markets/resource_files/maps/2013/new_york/ny_nmtc_pro jects.pdf. Novogradac & Company LLP. “Resources: Reports & Research.” Date accessed December 6, 2013. http://www.novoco.com/new_markets/resources/reports.php Poko Partners LLC. Portfolio of Properties. New York: Poko Partners LLC. http://www.pokopartners.com/brochure2/. Porter, Michael. The Competitive Advantage of the Inner City. Harvard Business Review (1995) May-June:55-71. Ridder, Geert. “Lecture 11: Differences-in-differences.”Lecture, Economics 513 from University of Southern California, Los Angeles, April 2012. Russell, J.R and Thomas G. Noland. “An Analysis of New Market Tax Credits.” Real Estate Issues 36, no. 3 (2011): 27-36. Studenmund, A.H. Using Econometrics: A Practical Guide. 5th ed. Boston: Pearson Education Inc, 2006. Trufund Financial Services. “Empowerment Reinvestment Fund LLC (ERF).” Date accessed December 6, 2013. http://www.trufund.org/erf/. United Nations. World Urbanization Prospects The 2011 Revision. New York: United Nations, 2012. United States Census Bureau. “2014 American Community Survey form.” Date accessed January 3, 2013. http://www.census.gov/acs/www/Downloads/questionnaires/2014/Quest14.pdf. United States Census Bureau. “When to use 1-year, 3-year, or 5-year estimates.” Last modified September 12, 2013. http://www.census.gov/acs/www/guidance_for_data_users/estimates/ United States Census Bureau. “How the Census Bureau Measures Poverty.” Last modified December 3,2013. http://www.census.gov/hhes/www/poverty/about/overview/measure.html United States Census Bureau. “American Factfinder.”Date accessed December 21, 2013. http://factfinder2.census.gov/faces/nav/jsf/pages/searchresults.xhtml?refresh=t. United States Department of Housing and Development. Community Development Block Grant Program – CDBG. Accessed December 1, 2013. http://portal.hud.gov/hudportal/HUD?src=/program_offices/comm_planning/communityd evelopment/rulesandregs/laws/sec5318 United States Department of Housing and Development. Sec. 5318.* Urban development action grants [* Section 118 of the Act. Accessed December 1, 2013. 67 http://portal.hud.gov/hudportal/HUD?src=/program_offices/comm_planning/communityd evelopment/rulesandregs/laws/sec5318 Unites States Government Accountability Office, Tax Policy: New Markets Tax Credit Appears to Increase Investment by Investors in Low-Income Communities, but Opportunities Exist to Better Monitor Compliance. GAO-07-296.Washington, DC (2007). Unites States Government Accountability Office, New Markets Tax Credit: Minority Entities Are Less Successful in Obtaining Awards Than Non-Minority Entities. GAO-09- 536.Washington, DC (2009). Unites States Government Accountability Office, New Markets Tax Credit: The Credit Help Fund a Variety of Projects in Low-Income Communities, but Could be Simplified .GAO- 10-334.Washington, DC (2010). Unites States Government Accountability Office, Community Development Financial Institutions and New Markets Tax Credit Programs in Metropolitan and Nonmetropolitan Areas. GAO-12-547R.Washington, DC (2012). United States Senate. Public Works and Economic Development Act of 1965. Last Accessed December 1, 2013. http://www.epw.senate.gov/pweda65.pdf White, Sammis B., Richard D. Bingham, and Edward W. Hill,eds. Financing Economic Development in the 21st Century. New York: M.E. Sharpe, 2003. Weintraub, E. “Neoclassical Economics.” Date accessed December 1. 2013. http://www.econlib.org/library/Enc1/NeoclassicalEconomics.html Wooldridge, Jeffrey M. Econometrics Analysis of Cross Section and Panel Data. 2nd ed. Cambridge: Massachusetts Institute of Technology, 2010. 68 APPENDIX American Community Survey Sample Questionnaire sourced from the U.S. Census Bureau 69 70 71 72 73 74 75 76 2012 U.S. Census Bureau Poverty Thresholds STATA Log File name: <unnamed> log: C:\Users\JJ\Documents\ThesisFinal.smcl log type: smcl opened on: 02 Feb 2014, 00:02:34 . import excel "C:\Users\JJ\Dropbox\THESIS\FinaldatasetV1.xlsx", sheet("Regions") firstrow . ttest PvtyBaseline, by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 26.38979 .3074466 12.78032 25.78678 26.99279 1 | 116 32.83519 1.125372 12.12062 30.60604 35.06433 ---------+-------------------------------------------------------------------- combined | 1844 26.79524 .2988346 12.8325 26.20915 27.38133 ---------+-------------------------------------------------------------------- diff | -6.445403 1.22195 -8.841956 -4.04885 77 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = -5.2747 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.0000 Pr(|T| > |t|) = 0.0000 Pr(T > t) = 1.0000 . ttest Pvty , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 25.52674 .3257017 13.53917 24.88792 26.16555 1 | 116 28.23017 1.214708 13.0828 25.82407 30.63627 ---------+-------------------------------------------------------------------- combined | 1844 25.6968 .3149239 13.52341 25.07916 26.31445 ---------+-------------------------------------------------------------------- diff | -2.703436 1.295899 -5.245021 -.1618513 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = -2.0861 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.0186 Pr(|T| > |t|) = 0.0371 Pr(T > t) = 0.9814 . ttest pvtych , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 -.863049 .2521045 10.47979 -1.357511 -.3685868 1 | 116 -4.605016 1.138714 12.26432 -6.860588 -2.349443 ---------+-------------------------------------------------------------------- combined | 1844 -1.098444 .2476843 10.63601 -1.584215 -.6126726 ---------+-------------------------------------------------------------------- diff | 3.741967 1.016682 1.747996 5.735937 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = 3.6806 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.9999 Pr(|T| > |t|) = 0.0002 Pr(T > t) = 0.0001 . ttest UnemplBaseline , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 78 0 | 1728 6.546038 .1200639 4.990962 6.310552 6.781524 1 | 116 6.971185 .353541 3.807753 6.270888 7.671481 ---------+-------------------------------------------------------------------- combined | 1844 6.572783 .1146939 4.925163 6.347839 6.797726 ---------+-------------------------------------------------------------------- diff | -.4251464 .4724139 -1.351669 .5013766 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = -0.8999 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.1841 Pr(|T| > |t|) = 0.3683 Pr(T > t) = 0.8159 . ttest LessHSBaseline , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 33.03441 .321085 13.34725 32.40466 33.66417 1 | 116 37.35568 1.207755 13.00792 34.96336 39.74801 ---------+-------------------------------------------------------------------- combined | 1844 33.30625 .3112112 13.36397 32.69589 33.91661 ---------+-------------------------------------------------------------------- diff | -4.321271 1.278173 -6.828092 -1.81445 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = -3.3808 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.0004 Pr(|T| > |t|) = 0.0007 Pr(T > t) = 0.9996 . ttest BachlorsBaseline , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 15.16339 .2647463 11.0053 14.64413 15.68265 1 | 116 15.18373 1.168336 12.58337 12.86948 17.49798 ---------+-------------------------------------------------------------------- combined | 1844 15.16467 .2586609 11.10737 14.65737 15.67197 ---------+-------------------------------------------------------------------- diff | -.0203415 1.065635 -2.110322 2.069638 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = -0.0191 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.4924 Pr(|T| > |t|) = 0.9848 Pr(T > t) = 0.5076 79 . ttest AfricanAmericanBaseline , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 26.92105 .7146584 29.70779 25.51937 28.32274 1 | 116 37.37674 2.669841 28.75506 32.0883 42.66518 ---------+-------------------------------------------------------------------- combined | 1844 27.57879 .6927915 29.74973 26.22005 28.93753 ---------+-------------------------------------------------------------------- diff | -10.45569 2.843758 -16.03302 -4.878361 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = -3.6767 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.0001 Pr(|T| > |t|) = 0.0002 Pr(T > t) = 0.9999 . ttest Under18baseline , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 27.31706 .1848575 7.684384 26.9545 27.67963 1 | 116 26.96797 .8570355 9.230555 25.27035 28.6656 ---------+-------------------------------------------------------------------- combined | 1844 27.2951 .1813675 7.788249 26.9394 27.65081 ---------+-------------------------------------------------------------------- diff | .349091 .7471565 -1.116272 1.814454 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = 0.4672 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.6798 Pr(|T| > |t|) = 0.6404 Pr(T > t) = 0.3202 . ttest baseline65 , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 11.34061 .1469854 6.110069 11.05233 11.6289 1 | 116 10.8296 .6093965 6.563401 9.622498 12.03669 ---------+-------------------------------------------------------------------- combined | 1844 11.30847 .1429595 6.13894 11.02809 11.58885 ---------+-------------------------------------------------------------------- diff | .5110181 .5888464 -.6438586 1.665895 ------------------------------------------------------------------------------ 80 diff = mean(0) - mean(1) t = 0.8678 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.8072 Pr(|T| > |t|) = 0.3856 Pr(T > t) = 0.1928 . ttest Migration2000 , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 .2235231 .0046825 .194647 .2143392 .2327071 1 | 116 .2667804 .0175993 .1895505 .2319195 .3016413 ---------+-------------------------------------------------------------------- combined | 1844 .2262443 .0045309 .1945638 .2173581 .2351305 ---------+-------------------------------------------------------------------- diff | -.0432572 .0186391 -.0798133 -.0067012 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = -2.3208 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.0102 Pr(|T| > |t|) = 0.0204 Pr(T > t) = 0.9898 . generate migrationch = Migration2012 - Migration2000 . generate unemplch = Unempl - UnemplBaseline . ttest Unempl , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 7.115104 .0886781 3.686278 6.941176 7.289032 1 | 116 6.99569 .2951974 3.179374 6.41096 7.580419 ---------+-------------------------------------------------------------------- combined | 1844 7.107592 .0851341 3.655811 6.940623 7.274562 ---------+-------------------------------------------------------------------- diff | .1194145 .3507257 -.5684471 .8072762 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = 0.3405 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.6332 Pr(|T| > |t|) = 0.7335 Pr(T > t) = 0.3668 . ttest LessHS , by(Treatment) Two-sample t test with equal variances 81 ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 23.78204 .2969241 12.34291 23.19967 24.36441 1 | 116 24.59855 1.201529 12.94087 22.21855 26.97855 ---------+-------------------------------------------------------------------- combined | 1844 23.83341 .288281 12.37931 23.26801 24.3988 ---------+-------------------------------------------------------------------- diff | -.816511 1.187513 -3.145523 1.512501 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = -0.6876 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.2459 Pr(|T| > |t|) = 0.4918 Pr(T > t) = 0.7541 . ttest Bachelors , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 20.92564 .3307957 13.75092 20.27683 21.57444 1 | 116 25.63448 1.638743 17.6498 22.38845 28.88052 ---------+-------------------------------------------------------------------- combined | 1844 21.22185 .3276256 14.06884 20.5793 21.86441 ---------+-------------------------------------------------------------------- diff | -4.708846 1.345291 -7.347302 -2.07039 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = -3.5002 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.0002 Pr(|T| > |t|) = 0.0005 Pr(T > t) = 0.9998 . ttest AfricanAmerican , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 25.68108 .6786801 28.2122 24.34995 27.0122 1 | 116 32.19224 2.419838 26.06246 27.39901 36.98548 ---------+-------------------------------------------------------------------- combined | 1844 26.09067 .6548321 28.11968 24.80638 27.37496 ---------+-------------------------------------------------------------------- diff | -6.511165 2.693519 -11.79384 -1.228493 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = -2.4173 Ho: diff = 0 degrees of freedom = 1842 82 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.0079 Pr(|T| > |t|) = 0.0157 Pr(T > t) = 0.9921 . ttest Under18 , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 23.94644 .1919935 7.981019 23.56988 24.32301 1 | 116 22.40036 .768453 8.276492 20.8782 23.92252 ---------+-------------------------------------------------------------------- combined | 1844 23.84918 .1864482 8.006425 23.48351 24.21485 ---------+-------------------------------------------------------------------- diff | 1.546082 .7672872 .0412375 3.050926 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = 2.0150 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.9780 Pr(|T| > |t|) = 0.0440 Pr(T > t) = 0.0220 . ttest Over65 , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 11.48495 .1412888 5.873265 11.20784 11.76207 1 | 116 12.0819 .7361012 7.928053 10.62382 13.53997 ---------+-------------------------------------------------------------------- combined | 1844 11.52251 .1402413 6.022216 11.24746 11.79755 ---------+-------------------------------------------------------------------- diff | -.5969428 .5776009 -1.729764 .5358786 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = -1.0335 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.1508 Pr(|T| > |t|) = 0.3015 Pr(T > t) = 0.8492 . ttest Migration2012 , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 5.080592 .1425895 5.927335 4.800926 5.360259 1 | 116 6.445337 .5428664 5.84685 5.370023 7.520651 ---------+-------------------------------------------------------------------- combined | 1844 5.166444 .1380939 5.930005 4.895607 5.437281 83 ---------+-------------------------------------------------------------------- diff | -1.364744 .5680324 -2.478799 -.2506895 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = -2.4026 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.0082 Pr(|T| > |t|) = 0.0164 Pr(T > t) = 0.9918 . ttest migrationch , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 4.857069 .1398972 5.815417 4.582683 5.131455 1 | 116 6.178556 .5345199 5.756956 5.119775 7.237337 ---------+-------------------------------------------------------------------- combined | 1844 4.9402 .1355104 5.819064 4.674429 5.20597 ---------+-------------------------------------------------------------------- diff | -1.321487 .5574284 -2.414745 -.2282293 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = -2.3707 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.0089 Pr(|T| > |t|) = 0.0179 Pr(T > t) = 0.9911 . ttest unemplch , by(Treatment) Two-sample t test with equal variances ------------------------------------------------------------------------------ Group | Obs Mean Std. Err. Std. Dev. [95% Conf. Interval] ---------+-------------------------------------------------------------------- 0 | 1728 .5690659 .1339663 5.568876 .3063126 .8318193 1 | 116 .024505 .3706535 3.99206 -.7096882 .7586983 ---------+-------------------------------------------------------------------- combined | 1844 .5348094 .1277036 5.483825 .2843504 .7852683 ---------+-------------------------------------------------------------------- diff | .5445609 .5259624 -.4869844 1.576106 ------------------------------------------------------------------------------ diff = mean(0) - mean(1) t = 1.0354 Ho: diff = 0 degrees of freedom = 1842 Ha: diff < 0 Ha: diff != 0 Ha: diff > 0 Pr(T < t) = 0.8497 Pr(|T| > |t|) = 0.3006 Pr(T > t) = 0.1503 . regress pvtych Treatment Source | SS df MS Number of obs = 1844 -------------+------------------------------ F( 1, 1842) = 13.55 84 Model | 1522.09095 1 1522.09095 Prob > F = 0.0002 Residual | 206966.918 1842 112.359891 R-squared = 0.0073 -------------+------------------------------ Adj R-squared = 0.0068 Total | 208489.009 1843 113.124802 Root MSE = 10.6 ------------------------------------------------------------------------------ pvtych | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- Treatment | -3.741967 1.016682 -3.68 0.000 -5.735937 -1.747996 _cons | -.863049 .2549962 -3.38 0.001 -1.363161 -.362937 ------------------------------------------------------------------------------ . regress pvtych Treatment Northwest Southwest Northeast Source | SS df MS Number of obs = 1844 -------------+------------------------------ F( 4, 1839) = 41.53 Model | 17274.416 4 4318.60399 Prob > F = 0.0000 Residual | 191214.593 1839 103.977484 R-squared = 0.0829 -------------+------------------------------ Adj R-squared = 0.0809 Total | 208489.009 1843 113.124802 Root MSE = 10.197 ------------------------------------------------------------------------------ pvtych | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- Treatment | -3.115278 .9795819 -3.18 0.001 -5.036488 -1.194069 Northwest | 6.976086 .870861 8.01 0.000 5.268106 8.684066 Southwest | 6.30611 .6673689 9.45 0.000 4.99723 7.614991 Northeast | 6.269086 1.16128 5.40 0.000 3.99152 8.546652 _cons | -2.731291 .2886228 -9.46 0.000 -3.297353 -2.165228 ------------------------------------------------------------------------------ . regress pvtych Treatment PvtyBaseline Source | SS df MS Number of obs = 1844 -------------+------------------------------ F( 2, 1841) = 128.84 Model | 25598.558 2 12799.279 Prob > F = 0.0000 Residual | 182890.451 1841 99.3429937 R-squared = 0.1228 -------------+------------------------------ Adj R-squared = 0.1218 Total | 208489.009 1843 113.124802 Root MSE = 9.9671 ------------------------------------------------------------------------------ pvtych | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- Treatment | -1.912907 .9631714 -1.99 0.047 -3.80193 -.0238841 PvtyBaseline | -.2837773 .0182285 -15.57 0.000 -.319528 -.2480267 _cons | 6.625774 .5374891 12.33 0.000 5.571622 7.679926 ------------------------------------------------------------------------------ . regress pvtych Treatment Northwest Southwest Northeast PvtyBaseline Source | SS df MS Number of obs = 1844 85 -------------+------------------------------ F( 5, 1838) = 76.31 Model | 35841.8137 5 7168.36274 Prob > F = 0.0000 Residual | 172647.196 1838 93.9320977 R-squared = 0.1719 -------------+------------------------------ Adj R-squared = 0.1697 Total | 208489.009 1843 113.124802 Root MSE = 9.6919 ------------------------------------------------------------------------------ pvtych | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- Treatment | -1.634486 .9369993 -1.74 0.081 -3.472181 .2032089 Northwest | 6.434774 .8286202 7.77 0.000 4.809638 8.05991 Southwest | 5.089504 .6401878 7.95 0.000 3.833932 6.345075 Northeast | 2.684439 1.132824 2.37 0.018 .4626809 4.906196 PvtyBaseline | -.2573434 .0183039 -14.06 0.000 -.2932421 -.2214447 _cons | 4.462827 .5805896 7.69 0.000 3.324142 5.601511 ------------------------------------------------------------------------------ . regress unemplch Treatment Source | SS df MS Number of obs = 1844 -------------+------------------------------ F( 1, 1842) = 1.07 Model | 32.2354451 1 32.2354451 Prob > F = 0.3006 Residual | 55391.0865 1842 30.0711653 R-squared = 0.0006 -------------+------------------------------ Adj R-squared = 0.0000 Total | 55423.3219 1843 30.0723396 Root MSE = 5.4837 ------------------------------------------------------------------------------ unemplch | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- Treatment | -.5445609 .5259624 -1.04 0.301 -1.576106 .4869844 _cons | .5690659 .1319178 4.31 0.000 .3103418 .82779 ------------------------------------------------------------------------------ . regress unemplch Treatment Northwest Southwest Northeast Source | SS df MS Number of obs = 1844 -------------+------------------------------ F( 4, 1839) = 2.58 Model | 308.797528 4 77.199382 Prob > F = 0.0360 Residual | 55114.5244 1839 29.9698338 R-squared = 0.0056 -------------+------------------------------ Adj R-squared = 0.0034 Total | 55423.3219 1843 30.0723396 Root MSE = 5.4745 ------------------------------------------------------------------------------ unemplch | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- Treatment | -.4596849 .5259122 -0.87 0.382 -1.491133 .571763 Northwest | .9111465 .4675428 1.95 0.051 -.0058241 1.828117 Southwest | .6301186 .3582932 1.76 0.079 -.0725856 1.332823 Northeast | 1.282291 .6234613 2.06 0.040 .0595247 2.505057 _cons | .3340591 .1549541 2.16 0.031 .0301546 .6379636 ------------------------------------------------------------------------------ 86 . regress Treatment UnemplBaseline Source | SS df MS Number of obs = 1844 -------------+------------------------------ F( 1, 1842) = 0.81 Model | .047774039 1 .047774039 Prob > F = 0.3683 Residual | 108.655046 1842 .058987539 R-squared = 0.0004 -------------+------------------------------ Adj R-squared = -0.0001 Total | 108.70282 1843 .058981454 Root MSE = .24287 ------------------------------------------------------------------------------ Treatment | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- UnemplBase~e | .0010337 .0011487 0.90 0.368 -.0012191 .0032866 _cons | .0561122 .0094335 5.95 0.000 .0376107 .0746137 ------------------------------------------------------------------------------ . regress unemplch Treatment Northwest Southwest Northeast UnemplBaseline Source | SS df MS Number of obs = 1844 -------------+------------------------------ F( 5, 1838) = 502.50 Model | 32007.9563 5 6401.59125 Prob > F = 0.0000 Residual | 23415.3657 1838 12.7395896 R-squared = 0.5775 -------------+------------------------------ Adj R-squared = 0.5764 Total | 55423.3219 1843 30.0723396 Root MSE = 3.5693 ------------------------------------------------------------------------------ unemplch | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- Treatment | -.219129 .3429192 -0.64 0.523 -.8916812 .4534231 Northwest | .5097503 .3049356 1.67 0.095 -.0883064 1.107807 Southwest | -.3126372 .234364 -1.33 0.182 -.7722848 .1470104 Northeast | -.9074041 .4088489 -2.22 0.027 -1.709261 -.105547 UnemplBase~e | -.84923 .0170247 -49.88 0.000 -.8826197 -.8158402 _cons | 6.176604 .1546777 39.93 0.000 5.873242 6.479967 ------------------------------------------------------------------------------ . regress migrationch Treatment Source | SS df MS Number of obs = 1844 -------------+------------------------------ F( 1, 1842) = 5.62 Model | 189.830828 1 189.830828 Prob > F = 0.0179 Residual | 62216.9332 1842 33.7768367 R-squared = 0.0030 -------------+------------------------------ Adj R-squared = 0.0025 Total | 62406.764 1843 33.8615106 Root MSE = 5.8118 ------------------------------------------------------------------------------ migrationch | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- Treatment | 1.321487 .5574284 2.37 0.018 .2282293 2.414745 _cons | 4.857069 .1398098 34.74 0.000 4.582867 5.131271 87 ------------------------------------------------------------------------------ . regress migrationch Treatment Northwest Southwest Northeast Source | SS df MS Number of obs = 1844 -------------+------------------------------ F( 4, 1839) = 5.18 Model | 694.766136 4 173.691534 Prob > F = 0.0004 Residual | 61711.9979 1839 33.557367 R-squared = 0.0111 -------------+------------------------------ Adj R-squared = 0.0090 Total | 62406.764 1843 33.8615106 Root MSE = 5.7929 ------------------------------------------------------------------------------ migrationch | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- Treatment | 1.440076 .5564999 2.59 0.010 .3486378 2.531514 Northwest | .3558933 .4947356 0.72 0.472 -.6144093 1.326196 Southwest | 1.227252 .3791319 3.24 0.001 .4836775 1.970826 Northeast | 1.61927 .6597224 2.45 0.014 .3253862 2.913154 _cons | 4.559061 .1639664 27.80 0.000 4.237481 4.880641 ------------------------------------------------------------------------------ . regress migrationch Treatment Migration2000 Source | SS df MS Number of obs = 1844 -------------+------------------------------ F( 2, 1841) = 419.41 Model | 19534.3151 2 9767.15754 Prob > F = 0.0000 Residual | 42872.449 1841 23.2875877 R-squared = 0.3130 -------------+------------------------------ Adj R-squared = 0.3123 Total | 62406.764 1843 33.8615106 Root MSE = 4.8257 ------------------------------------------------------------------------------- migrationch | Coef. Std. Err. t P>|t| [95% Conf. Interval] --------------+---------------------------------------------------------------- Treatment | .6001363 .4635279 1.29 0.196 -.3089594 1.509232 Migration2000 | 16.67584 .5785908 28.82 0.000 15.54108 17.81061 _cons | 1.129632 .1737885 6.50 0.000 .7887886 1.470475 ------------------------------------------------------------------------------- . regress migrationch Treatment Northwest Southwest Northeast Migration2000 Source | SS df MS Number of obs = 1844 -------------+------------------------------ F( 5, 1838) = 167.73 Model | 19552.9921 5 3910.59843 Prob > F = 0.0000 Residual | 42853.7719 1838 23.3154363 R-squared = 0.3133 -------------+------------------------------ Adj R-squared = 0.3114 Total | 62406.764 1843 33.8615106 Root MSE = 4.8286 ------------------------------------------------------------------------------- migrationch | Coef. Std. Err. t P>|t| [95% Conf. Interval] --------------+---------------------------------------------------------------- Treatment | .616123 .4647702 1.33 0.185 -.2954102 1.527656 88 Northwest | .3129692 .412386 0.76 0.448 -.4958251 1.121763 Southwest | .0344355 .3187936 0.11 0.914 -.5908003 .6596712 Northeast | .2927792 .5518813 0.53 0.596 -.789601 1.375159 Migration2000 | 16.65402 .5855856 28.44 0.000 15.50554 17.8025 _cons | 1.089274 .1832061 5.95 0.000 .7299596 1.448588 ------------------------------------------------------------------------------- . regress pvtych Treatment Northwest Southwest Northeast Southeast, noconstant Source | SS df MS Number of obs = 1844 -------------+------------------------------ F( 5, 1839) = 37.51 Model | 19499.3477 5 3899.86953 Prob > F = 0.0000 Residual | 191214.593 1839 103.977484 R-squared = 0.0925 -------------+------------------------------ Adj R-squared = 0.0901 Total | 210713.941 1844 114.270033 Root MSE = 10.197 ------------------------------------------------------------------------------ pvtych | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- Treatment | -3.115278 .9795819 -3.18 0.001 -5.036488 -1.194069 Northwest | 4.244795 .825963 5.14 0.000 2.624872 5.864719 Southwest | 3.57482 .6064908 5.89 0.000 2.385337 4.764303 Northeast | 3.537796 1.126317 3.14 0.002 1.328802 5.746789 Southeast | -2.731291 .2886228 -9.46 0.000 -3.297353 -2.165228 ------------------------------------------------------------------------------ . tab1 Northwest Southwest Northeast Southeast, subpop(Treatment) -> tabulation of Northwest Northwest | Freq. Percent Cum. ------------+----------------------------------- 0 | 108 93.10 93.10 1 | 8 6.90 100.00 ------------+----------------------------------- Total | 116 100.00 -> tabulation of Southwest Southwest | Freq. Percent Cum. ------------+----------------------------------- 0 | 104 89.66 89.66 1 | 12 10.34 100.00 ------------+----------------------------------- Total | 116 100.00 -> tabulation of Northeast Northeast | Freq. Percent Cum. ------------+----------------------------------- 0 | 114 98.28 98.28 89 1 | 2 1.72 100.00 ------------+----------------------------------- Total | 116 100.00 -> tabulation of Southeast Southeast | Freq. Percent Cum. ------------+----------------------------------- 0 | 22 18.97 18.97 1 | 94 81.03 100.00 ------------+----------------------------------- Total | 116 100.00 . regress pvtych Treatment Northwest Southwest Northeast Southeast, noconstant Source | SS df MS Number of obs = 1844 -------------+------------------------------ F( 5, 1839) = 37.51 Model | 19499.3477 5 3899.86953 Prob > F = 0.0000 Residual | 191214.593 1839 103.977484 R-squared = 0.0925 -------------+------------------------------ Adj R-squared = 0.0901 Total | 210713.941 1844 114.270033 Root MSE = 10.197 ------------------------------------------------------------------------------ pvtych | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- Treatment | -3.115278 .9795819 -3.18 0.001 -5.036488 -1.194069 Northwest | 4.244795 .825963 5.14 0.000 2.624872 5.864719 Southwest | 3.57482 .6064908 5.89 0.000 2.385337 4.764303 Northeast | 3.537796 1.126317 3.14 0.002 1.328802 5.746789 Southeast | -2.731291 .2886228 -9.46 0.000 -3.297353 -2.165228 ------------------------------------------------------------------------------ . predict res, r . export excel using "correctedresiduals", firstrow(variables) file correctedresiduals.xls saved . bootstrap, reps(5000) cluster(Northwest Southwest Northeast Southeast Treatment > ) : regress pvtych Treatment Northwest Southwest Northeast Southeast, noconstan > t (running regress on estimation sample) Bootstrap replications (5000) ----+--- 1 ---+--- 2 ---+--- 3 ---+--- 4 ---+--- 5 x..xx..xx..xxxx........xx...x...........x.x...x... 50 x..xxx..x..x....x.x.x...x.x..xxxxxx......x..x.xx.x 100 xx....x..xx..x..xx..xxx.x..x.x.xx...xxx.x.x.....xx 150 ..x.x.x.......xx.xxx..xx..x...x.x.x..x.x..x.xxxx.. 200 .x..x.xx..x.x.xxxx.xx.x........xxxx...x.x..x...... 250 xx.x.....xxx..x..x.......xx.xx...xx.xx.x.x..x.x..x 300 90 x..x.x.x...x.x....x..x....x..x.x..x.xxxx.......... 350 xxx.....xx...x..x....xx.x..xx.xx.x....x...x..xx... 400 ...x.x..x........xx.x...xx..x.x...x...xxx...x.x.xx 450 x..xx.xx..x...x..x.x.....xxx.xx..xx.xxx.........x. 500 ..xx..xxx..xxxx......x.x.xxxx.x...x.xx.xxxxxx.xxxx 550 .x..x.x.x.x...xxx....x....xx.xx.....x...x..x...... 600 ..xx.x......xx...xx.x.xxx...x.x...x.xxx.x...xx.xx. 650 ..x...x.xx..x.xx..xx..x..xx.x..x..xxx.xx.x..xx...x 700 xx............x.xx...xxx..x.x.....x..x.x.xxxx.xx.x 750 xx..x.xxx.x.x...x...x..xxxxxx...xx...x..x..x.....x 800 ....x.x.x.x...x..x...xx.x..x....xx..x...xxxx..x... 850 ...xx..xxxxxxx.xx...xxxx....x..x.....xx.xx......x. 900 xx.xx.x.xx.xx..x.x.xxx.x.xx.xx.xx..............xx. 950 .............x...xx..xxxx...xxx....x.x.x.x.xx..x.. 1000 ....xx..x...x..x.x...xx.x....x...xx.x..xx...x.x... 1050 .x......x.xx...x...xx.....x.xx..xx.x.x..x..xx.x... 1100 .x.xxxxx....xx..xx.x.....x..xxxx.xx..x..x.x...x.xx 1150 ...x.x...........x.xx...x..xxxx.xxx..x...x.x...x.. 1200 xxx..x.xx..x..xxx.x.x..x.x..xx..x..x.xxx.x.....x.. 1250 ......x..x.x.x.x..x...x...x.xxx..x..x..xxx..xxx.x. 1300 .xx..........x.xx.xx...xxxx.x..x..xx......xx.....x 1350 x..x..x..xx...x....x.x...x.x.xx....xxx.....xx...xx 1400 .x.x.xxx...xx..xx....x.xxxx.......x..xx...x.x..xxx 1450 .x.xxx..x..x.x...x..xx.x.xxxx...x......x.xx.xxx... 1500 ...xx.x.xxx..x..x........xx.xxx.x..xx...x.x.x...x. 1550 xx.xxxxxxxxx.x..x......xxxx.xxx..x..x..xx.x..xxxx. 1600 x.xx..xxxxxx.xx....x..x........x...xx....xxx.x.x.. 1650 x.xx.xx.....xxx...xx..xx...x....x.x.xx..x.x..x..xx 1700 x..xx.x...x...xx..x.xxxxx..x.x.x..xx.xx....xxxx..x 1750 xx.x....x...xx.x.x.x..x....x..x..xxx..x...x..x..xx 1800 x.xx.xxx...xxxxx.xx.xx....xx....x......x...x...xxx 1850 xxxxxx.......x..x..x......x...xxxx....x........... 1900 ....x...xxx....x.....x.xxxx..xxxxx..xx..x...x..x.. 1950 x.x..xx.x..xx.x.xxx..xx..xx....x..xxx.xxx.xxxx..xx 2000 ....x..x..x..xx..xx...xx....xx.xx.xx....xx...x.xx. 2050 x.x..x...x.xx.xx.....xxx.xxxx.xxx...x..xx.x....... 2100 x.xx.x...x..xx...xxx.x....x...x.....xxx.xx.x.....x 2150 ..x.x...x.....xx.x.xxx....xx.x..x.xxxx...x...xxx.x 2200 .........x..x....x.....xx..x...x.xx..xx.xxxx.x..x. 2250 ...x...x....x...........x....xx.x..xx..xx..xxx.x.. 2300 ..xxx..xx..xx..x.xx.xx..x...xxxx..xxx.xxxx..x.xx.x 2350 x.xx...x..x.x.x.x..x.x...x..x...x..........xx...x. 2400 .....xx.x..xxx.x........x.....x....x...x.......x.. 2450 x..xxx..xx...x....x....x.....x.x.x..xxxx...xx.x.x. 2500 .....x..x.xx...x.x.x...x...x...xx..x.....x....xx.x 2550 x..x....x.xx.x.x.x..xx......x...xxx...x.....x...x. 2600 ..xxxxx.x.xx..x.x.xx..x..x.......x.......x..x..x.x 2650 .xx...x...xxx.x.......xx.x.x.x.....x....xxx.xxx.xx 2700 .x..xx.....xxx.x...xxx....xxx..x.x..x..x...xx.x... 2750 xxx.....x....x.xxx...xxxx...xxxx.x......x.x.x..xxx 2800 .x.xx..x..x..x.......xxxx.x..x..x.x.x.xx...xx..x.x 2850 91 ...x..xx.x....x.x...xx.x.....xx..x.x.xx.xxxx..xx.. 2900 .xxx.xx.x..x.x....x...xxx.x....xxx..xxxx.xx.....xx 2950 ...xx.....xxx....x......xxx..xxxx.x..x..x.x.xxxx.. 3000 xxx..xxx..x.xx........xxx.x....x.x.xxx..x.....xx.. 3050 .xx..xx....x.....xx..x..x...x.x.x..xx.....x.xx.x.x 3100 x..x.xx.x.x..x.x.x..x..xx..x.x.xxxxx...xx.x...x.xx 3150 x......x..xx.xxxxx.x.xxx.xxx.xx..xx..x......x....x 3200 .....x..xx.xx.....x..xx.xx.x...x.x..xxx........x.x 3250 x......xx...x.x.x.x.....x.xx.xxxxx..xx.........xx. 3300 x.xxx..xxxx....xxxxx.....x.x.xxx..x..xx.x.xxx..x.x 3350 xxx.xxx.....x.xx....x....x..xx.x..x...xxxx.x.x..xx 3400 .x..x.x..x..x..xx.x.xx....x..x.....x...xx.x....x.. 3450 .xxx..xx.........x.x..x...x.x...xxxxxx.xxx.x..x... 3500 xx...x.xx....x..x.x....x..x..x..x....x.x..x.x....x 3550 ..xx.xx..xxx...xxx..xx......xxxx.......xxx.xxx.x.x 3600 xxx..........x..xxx.....xx...xxx...xx...x.....xx.x 3650 ..x...x...xx.xx..xx.xx.xxx.x...x.xx....xx..x..xx.. 3700 x..xxx.xxxx.xx.x.....x.....xx........x.........xxx 3750 .....xx....xx.x...x.x...x.x..x...x.xxxxxx...x.x... 3800 xx.xx....x.....x..xx.x.xxxxx.x...x....xxx....xx..x 3850 ..xx..x.xx.xx.xx..x.x..x.x.xx...x.xxxx.xxxx...x..x 3900 ...x..x...x..xx...x.....xx...xx...xxx.xxx.x.xxx..x 3950 ...x.x....xxxx...xx.x...xx.x..x..xxxx.xx..xxxx.... 4000 .xxxx.x....x.x...x..x..xx.x..x.x.x......x...xxx... 4050 .xx.xx..xxx.x.xx..xxxx....x.x..xx.xxx..x..x..xxx.. 4100 .x.xxx.....x.x.x.xxxx.xxxxx...x.x..x..xxx...xxxx.x 4150 x........x..x...x..x.......xxxx....x...x.x..xx..x. 4200 x.x..x.......x..x...x.x.x.....x....xx.x...x.x..x.. 4250 ...xx.....x....xxx..xx..xxx.x.xx..xxxx.......xxxx. 4300 ...xx.xxxx.xxx.x..x.........x...x.xx.x..xx...x.x.x 4350 x.x...x..xxxx..xx..xx......xx..xxxxx....x..x...xxx 4400 .xxx..x..xx.xxxxx.x.x..x..xx.xxx..x.x.x..x.....xx. 4450 xx.xxxx...xx...x.xx.xx..x.x.x.xx..x....xx.x..xx.xx 4500 .xx....x.x..x...xx....xxx.....x.xx.xx...x.x..x.x.. 4550 xxx.x...xx.....xxx.x...x..xx.xxxx..x..xxxxx..xx.xx 4600 x...xxxxxx.x......xxx.x.x.xx.x..xxx....x......x... 4650 ..x.xx..x.xx...x....xx...xx..xx...xx.x...x..x..x.x 4700 .x..x...x.....xxx.xxxxx.xxx.x.xx.x.....x...x...x.. 4750 ...x...x..xx..x...x..x.....x.....x.xx.x.xxx.x..x.x 4800 x.xx.xxx..x...x..x..xxxx...x.xx..x.xx..x....x..x.. 4850 xx..xxxxx.x..xxxx...x....x...x.xx..x.x.........xxx 4900 ..xxx.x.x...x.xx..xxx..xxxxx....x...xxx.x.xx..xx.. 4950 .x.x....x.xxx.x.xx.....xx.x....x...x.x.x.xxxxxx... 5000 Linear regression Number of obs = 1844 Replications = 2916 Wald chi2(5) = 20.88 Prob > chi2 = 0.0009 R-squared = 0.0925 Adj R-squared = 0.0901 Root MSE = 10.1969 92 (Replications based on 8 clusters in Northwest Southwest Northeast Southeast Trea > tment) ------------------------------------------------------------------------------ | Observed Bootstrap Normal-based pvtych | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- Treatment | -3.115278 2.282657 -1.36 0.172 -7.589203 1.358647 Northwest | 4.244795 1.976416 2.15 0.032 .3710908 8.1185 Southwest | 3.57482 1.726847 2.07 0.038 .1902609 6.959379 Northeast | 3.537796 3.485765 1.01 0.310 -3.294178 10.36977 Southeast | -2.731291 1.602626 -1.70 0.088 -5.87238 .4097984 ------------------------------------------------------------------------------ Note: one or more parameters could not be estimated in 2084 bootstrap replicates; standard-error estimates include only complete replications. . log close name: <unnamed> log: C:\Users\JJ\Documents\ThesisFinal.smcl log type: smcl closed on: 04 Mar 2014, 01:25:10 93 Map of New York Counties

Abstract (if available)

Abstract The Community Renewal Tax Relief Act of 2000 was enacted to spur development in economically distressed communities within the United States. One of provisions of this bill was to create tax incentives for investment in small businesses within these distressed areas. This provision led to the creation of the federal economic development program, New Market Tax Credit Program, which draws private investors to low income neighborhoods through the provision of federal tax credits. Originally, the Program was approved for 7 years and has been extended in subsequent years. In 2013, this program came into question of whether or not it should be extended once again and whether it was working. However, supporting documentation of the impact of the program is sparse. Since the implementation of this program in 2003, there have been gaps in its program assessment, which this paper attempts to fill. This paper adopts both case study and quantitative methods to analyze the impact of the New Market Tax Credit Program on census tracts within New York. A difference in difference regression model was used to compare changes in poverty rates between low income census tracts that have received investment due to the Program and those have not and estimate the effect of the Program. Using the case study method, projects funded using New Market Tax Credits showed small employment growth and greater access to health and human services in high need neighborhoods. The results of the econometric analysis show that there is a statistically significant negative correlation between receiving investment due to the Program and the change in poverty rate.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Three essays on housing demographics: depressed housing access amid crisis of housing shortage

Asset Metadata

Creator Berman, Jacqueline (author)

Core Title Increased access to capital: evaluation of the New Market Tax Credit Program in New York

School Dual Degree

Degree Master of Planning / Master of Arts

Degree Program Planning / Economics

Publication Date 04/21/2014

Defense Date 03/05/2014

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag low-income census tracts,New Market Tax Credit Program,New York,OAI-PMH Harvest,poverty rate

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Ridder, Geert (committee chair), David, Joel (committee member), Mitchell, Leonard (committee member)

Creator Email bermanj@usc.edu,jberma2@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-381217

Unique identifier UC11297399

Identifier etd-BermanJacq-2380.pdf (filename),usctheses-c3-381217 (legacy record id)

Legacy Identifier etd-BermanJacq-2380.pdf

Dmrecord 381217

Document Type Thesis

Format application/pdf (imt)

Rights Berman, Jacqueline

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA