THE IMPACT OF STATE R&D TAX CREDITS IN AN ECONOMIC DOWNTURN

THE IMPACT OF STATE R&D TAX CREDITS IN AN ECONOMIC DOWNTURN A Thesis submitted to the Faculty of the Graduate School of Arts and Sciences of Georgetown University in partial fulfillment of the requirements for the degree of Master of Public Policy in Public Policy By Amanda L. O Connor, B.A. Washington, DC April 15, 2011

Copyright 2011 by Amanda L. O Connor All Rights Reserved ii

THE IMPACT OF STATE R&D TAX CREDITS IN AN ECONOMIC DOWNTURN Amanda L. O Connor Thesis Advisor: Matthew Fleming ABSTRACT The United States is an innovation-based economy that relies on research and development (R&D) to fuel economic growth. R&D tax credits are used by both state and federal policymakers to encourage additional private investments into R&D which, in turn, fuel economic growth. While empirical analysis generally concludes that R&D tax credits have a positive impact on GDP growth, the impact of R&D tax credits on economic growth during a recession has not been examined. This thesis investigates the impact of state-based R&D tax credits on state GDP growth during the economic downturn of the 2007-2009 recession. Data from the Bureau of Economic Analysis, Bureau of Labor Statistics, National Science Foundation, and legislative statutes are examined to determine the impact of R&D tax credits through a recession via a multivariate regression analysis. The results suggest that in an economic recession, the presence of an R&D tax credit dampens the severity of an economic downturn when compared, ceteris paribus, to states without such tax credit. However, R&D tax credit itself should not be seen as a magic bullet that works in isolation. Many factors influence economic performance, and R&D tax credits represent just one policy tool influencing economic growth within a state. iii

The research and the writing of this thesis is dedicated to my friends, family, colleagues and advisors who helped along the way. Your assistance from walking my dog, understanding why I had to leave work, and answering my questions was invaluable and I could not have completed this without your support. Many Thanks, Amanada L. O Connor iv

TABLE OF CONTENTS Chapter 1: Introduction and Background 1 Introduction. 1 Background. 3 Literature Review 9 Chapter 2: Framework and Data. 13 Conceptual Framework and Hypothesis. 13 Data and Methods... 14 Chapter 3: Descriptive Statistics... 19 Chapter 4: Results 23 Regression Findings. 23 Thoughts for Future Research.. 29 Chapter 5: Conclusion.. 30 Policy Implications... 30 Conclusion 31 Appendix A: Alternative Models. 32 Appendix B: Per capita GDP.. 33 Works Cited..... 34 v

SUMMARY OF TABLES AND FIGURES Table 1: Summary of State-Level R&D Tax Credits... 6 Table 2: Summary of Data Acquisition... 15 Table 3: Variables used in explanatory model.... 16 Table 4: Descriptive Statistics..... 19 Table 5: Descriptive Statistic Comparison for States with and without Tax Credits.. 20 Figure 1: Comparison of per capita GDP in states with and without R&D Tax Credits 21 Table 6: Dependent Variable Descriptive Statistic Comparison for States with and without Credits in 2006 and in 2009.. 22 Figure 2: GDP Distribution 25 Table 7: Explanatory Models with lgdp as the Dependent Variable.... 26 vi

Chapter 1: Introduction and Background Introduction The United States is an innovation-based economy, one that relies on research and development (R&D) to fuel growth. Indeed, innovation plays a critical role in the development of a 21 st -century economy, driving long-run growth (Litan 2009). Reports from the Brookings Institute and the Information Technology and Innovation Foundation (ITIF) describe how the development of an innovation-based economy takes place. Robert Atkinson of ITIF depicts the transformation to an innovation-based economy via the evolution of technological innovation and increased economic prosperity (Atkinson 2006). Muro, of Brookings, examines the change in the economy, calling an innovation based economy a different type of growth model, not of the technological bubbles in the past, but focused on long-term productive investments in innovative industries; sustained investments of this type can help to rebuild a crippled economy (Muro 2010). For the purpose of this paper, an innovation-based economy will be defined using Muro s definition. Capturing the competitive edge from innovation is pertinent to both the United States as a whole, as well as individual states. In January 2008, Secretary of Commerce Carlos Gutierrez commissioned a report from the Advisory Committee on Measuring Innovation in the 21st Century Economy. The committee, established by the Secretary of Commerce, explored the state of innovation in the economy. First it defined innovation as the design, invention, development and/or implementation of new or altered products, services, processes, systems, organizational structures, or business models for the purpose of creating new value for customers and financial returns for the firm (Committee, 2008). The committee reiterated the importance 1

of innovation in the economy because innovation is directly linked to the performance of the economy (Committee, 2008). Policymakers on both the national and state level are focused on encouraging innovation as a way to grow the economy and gain a competitive edge in the marketplace. Government policy, from tax credits to education subsidies, is a critical driver of economic investment for innovation on a state and federal level (Wilson 2005). Policymakers have the ability to influence the state of innovation via many avenues, including educational policy, direct investments in R&D, and policies that encourage private investment into innovation-based industries. Berguland and Clarke (2000) examined the changing state of the economy and the impact to the states. In their findings, they write, The U.S. Economy is undergoing a dramatic transformation as the nation moves to an economy driven by technology industries and the application of technology in traditional industries. To compete in this new economy states must have an economic base of firms that constantly innovate and maximize the use of technology in the workplace. Also critical is a strong research and development base that can provide these technology-intensive companies with access to state of the art research, researchers, and research facilities. The new economy they speak of is focused on encouraging innovation using policy tools, such as the R&D tax credit, to encourage innovation. The purpose of an R&D tax credit is to encourage companies to invest in more R&D than they would normally do, increasing total R&D investment to the socially 2

optimal level (Hall 2001). Typically, firms make investment decisions based on commercial returns (e.g., finding treatments for a disease) as opposed to basic science (e.g., genome mapping). As a result, firms often under-invest in basic science because the findings from basic science do not have a direct commercial application. The R&D tax credit is designed to help correct for this externality and encourage additional investments. The R&D tax credit is one such tool that has demonstrated effectiveness for economic growth. Economic evidence has consistently been presented (more later in the paper) regarding the effectiveness of the R&D tax credit and returns to GDP. What has not been examined is how the presence of an R&D tax credit may buffer an economic downturn, like the 2007-2009 recession. Background The research and experimentation tax credit, commonly referred to as the R&D tax credit, was first implemented in the United States in 1981. It is a commonly used policy tool to encourage incremental investments in research and development by private industries. President Reagan s administration implemented the R&D tax credit as a way to stimulate business growth and invest in future capabilities of the economy. Its purpose is to subsidize incremental investments in R&D, bringing total R&D expenditures closer to the socially-optimal level. Much of the focus on innovation and economic growth is at the country level. Countries from Ireland and Germany to Japan and the United States have developed public policies to develop an R&D tax credit (Atkinson 2007). In the United States, the federal government often has an R&D tax credit as mandated by legislation. However, this credit 3

is not permanent and must be periodically renewed (Atkinson 2006). In addition to the federal credit, many individual states have set policies dedicated to driving economic growth through research and development and innovation clusters. State investments into innovation via R&D tax credits have demonstrated incremental economic growth much like the impact of the federal research and development tax credit, in states that have implemented R&D tax credits. These investments into R&D are for the purposes of building a competitive economic advantage over neighbors. Details of this exchange are provided later in the paper. The credit allows companies to deduct qualified research expenses, including wages, supplies, and contract research, for incremental investments in R&D. Currently, the federal government offers a 20% credit on incremental investments, year to year increases in wages, supplies and contract research. The Congressional Office of Technology assessed the economic impact of the R&D tax credit and found that for every dollar lost in tax revenue, the R&D tax credit produces a dollar increase in reported R&D spending. Other studies show a larger benefit. Wilson suggests an average benefit 1.7 times the cost of the R&D tax credit (Wilson 2005) The federal R&D tax credit, as currently designed and implemented, is available for qualified research and development expenditures. 1 The credit is composed of three main parts: 1 Qualified research expenditure is defined as research in the laboratory or for experimental purposes, undertaken for discovering information, technological in nature, application is intended to be useful in the development of a new or improved business component for the tax payer, whether carried on by the taxpayer or on behalf of the taxpayer by a third party. This is about 62-65 percent of all R&D spending (Hall 2001). 4

1 Wages - For employees performing qualified research and development activities. 2 Supplies - Used for research activities. 3 Contractors - 65 percent of amount paid to contractors to perform research activities (Ernst & Young 2008). While many states have varying laws that dictate how a state-based R&D tax credit is implemented, they are primarily based on of the design of the federal R&D tax credit. Many states have passed a state-based R&D tax credit to encourage R&D expenditures in that state. Minnesota was the first state to do so in 1982. Today, thirty-two states have R&D tax credits. State R&D tax credits are designed to increase incremental investments in innovative industries, encouraging job growth in high-paying sectors. While every state R&D tax credit is unique, the majority is modeled after the federal R&D tax credit, allowing companies to deduct qualified research expenditures at a certain rate (see Table 1; Wheeler 2007). States are implementing R&D tax credits to encourage investments in the state economy that can serve as a catalyst for economic growth (State Science and Technology 1997). 5

Table 1: Summary of State-Level R&D Tax Credits (adapted from Wilson 2006) 2 State Year Enacted Statutory Credit Rate State Year Statutory Enacted Credit Rate Arizona 1994 11.0% Missouri 1994 6.5% California 1987 15.0% Montana 1999 5.0% Connecticut 1993 6.0% Nebraska 2005 3.0% Delaware 2000 10.0% New Jersey 1994 10.0% Georgia 1998 10.0% North Carolina 1996 5.0% Hawaii 2000 20.0% North Dakota 1988 4.0% Idaho 2001 5.0% Ohio 2004 7.0% Illinois 1990 6.5% Oregon 1989 5.0% Indiana 1985 5.0% Pennsylvania 1997 10.0% Iowa 1985 6.5% Rhode Island 1994 16.9% Kansas 1988 6.5% South Carolina 2001 5.0% Louisiana 2003 8.0% Texas 2001 5.0% Maine 1996 5.0% Utah 1999 6.0% Maryland 2000 10.0% Vermont 2003 10.0% Massachusetts 1991 10.0% West Virginia 1986 10.0% Minnesota 1982 2.5% Wisconsin 1986 5.0% R&D tax credits were used because when firms invest in R&D for the purpose of commercialization, they do not keep all of the knowledge to themselves; it is shared within the scientific community, as a public good (Griffith 2000). These positive externalities have benefits beyond the firm and, therefore, government policymakers have encouraged additional investments through R&D tax credits. The interaction between the government 2 Examination of state-specific statutes suggests that no changes in the R&D tax credits for the states listed were identified between 2006 and 2009. 6

and private industries can drive economic growth in innovative industries because of the shared scientific knowledge (Hall 2001). As such, government policies have played an important role in determining the types of investments made by private industry. Traditionally, government-funded projects focus on basic research, rather than the development or commercialization of science. The findings of basic research are often used by firms across industries to promote the development and commercialization of science into innovation products. This, in turn, helps to grow the economy. Spending by industry for R&D activities has grown more quickly than spending by the federal government (an average annual rate of 5.4 percent versus 3.5 percent) and has exceeded federal funding since 1980 ; since 1990, private R&D investment growth has been 6.2 percent (CBO 2007). A consensus amongst economists is that basic research, research without a commercial objective, provides greater social than private returns (Hall 2003). As such, there is an underinvestment in basic research because the incentives do not yield the same returns as applied R&D investments for commercial purposes (Hall 2001). 3 Additionally, if basic R&D were solely conducted in the private sector, without incentives for additional investments, then companies may only invest in R&D that directly benefits their commercial purposes. Basic research is a critical component of R&D because it provides the foundation for innovation and economic growth, yet private organizations are 3 According to analysis by Hall (2001), the social return to R&D is greater that the private return to R&D, which leads to underinvestment in private R&D. This is due to several factors. Often it is difficult to evaluate and fund some types of research because it would mean revealing ideas which can benefit competitors. Additionally, large organizations are often bureaucratic and are best at commercialization, while smaller firms are better at innovation. Helping the smaller firms invest (when they do not have as much capital to being with) induces a cycle of innovation and commercialization. R&D tax credits (and other policies that encourage R&D) help to correct this imperfect market by changing the financial constraints under which firms operate, and account for societal externalities. 7

sometimes reluctant to invest in basic R&D because the findings would not be proprietary and financial rewards would be spread across firms that didn t invest the capital resources in research (Hall 1999). Not all government subsidies must come in the form of R&D tax credits. Both the federal and state governments directly invest in R&D, provide direct subsidies to supplement research, and provide a strong academic talent pool for employers to pull from. The impact of each of these policies varies. Direct funding usually increases overall R&D activity because direct funding of government research is often in the field of basic science which private firms can and do leverage for commercialization purposes. R&D tax credits are a way to supplement private R&D in addition to the activities undertaken by the government and academia. Government subsidies can reduce the amount of private R&D expenditures because firms would have invested in that research regardless of the subsidy and they can increase the cost of R&D by inducing additional competition. Finally, government investment in academia and education is a pivotal indicator of R&D investments. Firms tend to locate near academic centers, hire employees from these institutions, and build long-term collaborations (Wu 2000). The literature broadly suggests that state-based R&D tax credits can boost economic growth in that state. However, do the tax credits help to minimize the contraction of a state s economy in an economic downturn? What policies should be implemented to encourage investments into R&D? This paper examines the extent to which investments in innovation can dampen an economic downturn in a specific geography. 8

Literature Review An innovation-based economy requires several elements for success; these can be applied on both the state and federal level to encourage additional investment and the relocation of firms. An intellectual infrastructure and a strong partnership with academia are crucial to attracting firms (Muro 2010). According to Berguland, the Milken Institute found that of the top thirty R&D metropolitan areas, twenty-nine were within close proximity to a major university (Berguland 2000). This leads to knowledge spillover, the building of innovation clusters, and a technically skilled workforce. Lastly, the cost of capital, or the availability of venture capital, and the physical infrastructure within the state to encourage new investments in R&D are important economic incentives for firms to make investment decisions (Berglund 2000). A primary purpose of government subsidy in research and development is to encourage additional investment of R&D to the socially optimal level. Griffith (2000) presents an empirical framework to examine how the social rate of return is calculated. Her findings indicate that for a 10 percent increase in R&D activities, a rate of return of 41.7 percent should be expected in the United States. This is higher than the private rates of return are around 27 percent (2000). Economists have calculated this gap based on knowledge spillover and which firm is able to commercialize and/or patent the results of research. The R&D tax credit rate is not the only factor associated with the effectiveness of a credit. The longevity of the credit and the ability of firms to depend on its availability is a key component of R&D decisions, making an R&D tax credit permanent increases the 9

level of R&D investment by private firms (Atkinson 2006). An R&D tax credit should be quasi-permanent in order to obtain the full benefit to the incentive. The rationale behind this is that many firms will respond differently to long and short term changes in the costs of R&D (Hall 1999). Griffith s analysis also extends to the impact of R&D tax credits finding that a 10 percent decrease in the reduction in the price of R&D would lead to a 1 percent increase in the amount of R&D in the short run and a 10% increase in the long run (Griffith 2000). On a state level, the goal of R&D tax credits is often to increase total levels of R&D in a particular state. But the impact of a state credit does not just increase R&D, it influences the location of R&D as firms move to states with more favorable R&D policies, primarily tax credits (Atkinson 2007). Wilson of the Federal Reserve Bank of San Francisco has completed the most comprehensive analysis of the impact of state based R&D tax credits. He concluded that state R&D tax credit increase research and development within the state, however, it comes at the expense of other states. A 1 percent decrease in the after-tax costs of R&D leads to a 1.7 percent increase in state R&D (Wilson 2005). In examining the impact of a federal vs. state R&D tax credit, some economists argue that a state-based R&D tax credit is not as effective, because not all of the R&D projects can be contained within the state s borders, while a federal credit can fall under the umbrella of United States intellectual property law (Wu 2005). 4 However, investments in 4 Intellectual property laws cover nations and not individual states. In the United States, all economic activity is conducted under one set of intellectual property laws, which is one patent system, one enforcement system, etc. A federally-based R&D tax credit spurs investments in R&D no matter where in the United States a firm is located. A state-based R&D tax credit only benefits firms within that state. What often motivates the state- 10

education can distinctly encourage firms to invest in a particular state. Investments in K-12 science, technological, engineering, and mathematical literacy are a primary driver for postsecondary educational success and pathways to hiring firms. It benefits states to encourage this investment because employees in high-technology industries make significantly more than those in other industries ; this encourages investments within a particular state (Berguland and Clarke, 2000). Comprehensive research has been completed on the effectiveness of an R&D tax credit on a federal level. However, because of the varying nature of state-based R&D tax credits, the same level of comparative empirical analysis does not exist. Analysis on R&D tax credits is primarily focused on a comparative country-level economic impact v. stateby-state comparison. However, some quantitative analysis exists on the state level. Wu presents the most detailed analysis, finding that the presence of an R&D tax credit does have a positive impact on company R&D tax expenditure (Wu 2005). The mere presence of a credit generates an additional $75-120 per capita. Consistent with the findings of Wilson and Atkinson, Wu finds that state based R&D tax credits increase private investment of R&D and that firms relocate to take advantage of these credits (Wu 2005). The Office of Technology and Assessment presents similar findings, identifying that the tax credit produces at least one dollar of new R&D spending for each dollar lost in tax revenue (State Science and Technology Institute 1997). based credit is not pulling investments from other countries, but rather surrounding states. The benefit to the country is additional R&D investments while the direct benefit to the state is increased jobs and direct investments into the state economy where the R&D is taking place. 11

Concurrent with other findings, Hall finds that the greatest spillover of R&D comes from basic research. She continues to conclude that the goal for a state R&D tax credit isn t the same as the federal government, increasing overall R&D investments, but it is to localize R&D into innovation clusters by localizing spillover effects from government, academia, and industry, to attract new investment and build economic growth within a state (Hall 1999). In California, two distinct innovation clusters have appeared: a high-tech industry in Silicon Valley and a life sciences / biotechnology cluster in San Diego. With regard to California and the innovation cluster, Hall finds that encouraging firms to move to your state early in the development of a new industry will probably mean that other firms will be attracted in the future, and that other firms in the state are more likely to benefit from knowledge spillovers from the new industry because of their geographical proximity (Hall 1999). 12

Chapter 2: Framework and Hypothesis Conceptual Framework and Hypothesis My hypothesis is that state-based R&D tax credits not only increase economic investments in states, but that their presence is indicative of innovation-based industries and can help dampen an economic downturn. Although the economic recession that started in 2007 officially ended in 2009, there are signs that another dip in the economy may occur (Bureau of Economic Analysis 2011). My analysis focuses on economic, educational, and production indicators, testing to see if states with the presence of an R&D tax credit in 2007 suffered less of an economic downturn through the end of 2009, ceteris paribus. The model examines the change in GDP from 2007 to 2009 as a function of the presence of state based R&D tax credits. I will use economic indicators (GDP and unemployment), R&D variables (government/academic R&D investments, patents issues, science and math education, and capital expenditures), and R&D tax credit information to examine the effectiveness of these credits on dampening an economic downturn. My hypothesis is that unemployment rates, on average, were lower in states with innovation investments, including a state based R&D tax credit. Furthermore, GDP increase over time is higher in states with the presence of a state based R&D tax credit. Based on analysis the extant literature, I developed my conceptual framework. The primary framework was adopted from Berglund and Clarke s analysis of the primary elements of a technology-based economy. Berglund and Clarke (2000) highlight that there are several keys to this economy: intellectual infrastructure/ technically skilled workforce 13

(measured as educational attainment), spillovers of knowledge (partnerships with academia, measured by academic patents), and capital (measured by the total number of prospering firms; Berglund and Clarke 2000). The conceptual framework is as follows with each of the indicators/investments in innovation positively correlated with state GDP growth: GDP = Economic indicators (GDP per capita & unemployment) + Educational Attainment (number of people with bachelor degrees) + Productivity (academic and commercial patents) + Private investments + R&D credit information. Data and Methods The data that I utilize comes from four primary sources: Bureau of Economic Analysis, Bureau of Labor Statistics, National Science Foundation, and Legislative Statutes. The thesis examines the state-year impact of R&D tax credits from 1997-2009, primarily focusing on an economic recession occurred from 2007-2009. The data I examine includes economic indicators (GDP and unemployment rates), R&D-based investments (R&D expenditures, education measures, capital venture funds), plus legislative information regarding the presence of a state R&D tax credit see Table 1). 14

Table 2: Summary of Data Acquisition Source Data Set Variables Time Period Acquisition Method Economic Indicators Bureau of Economic Analysis Regional Economic GDP 1997-2009 Downloaded from Accounts bea.gov Economic Indicators Bureau of State Labor Statistics Unemployment Unemployment 1997-2009 Downloaded from bls.gov R&D Investments National Science Foundation Science and Engineering Indicators, State Indicators Education, R&D Investments, R&D Outputs 2007 Downloaded from nsf.gov State Credits State Legislation Read Legislation Credit, Rate, Time Varies (based on state law) Read legislation The analysis focuses on the impact of a state-based R&D tax credit on GDP, specifically during the 2007-2009 recession. The analysis examines changes based on a state-year basis going back to 1997, thereby allowing for a comparison period prior to the recession. The regression model is a multivariate regression, examining the impact of state based R&D tax credits while controlling for other factors often associated with an innovation based economy (such as education, employment, patents, venture capital). Variables for the model were picked based on previous research on R&D tax credits (See Table 2). Previous empirical research often controls for R&D inputs (e.g. funding), education, R&D outputs (e.g. patents issued), and the availability of funding. The model is based on a consensus within the literature on the elements that are critical for an innovation-based economy in which R&D tax credits can be successful. Berguland and 15

Clarke (2000) concisely define these elements as: intellectual infrastructure, spillovers of knowledge, physical infrastructure, technically skilled workers, and capital. To build a robust analysis, variables for these elements are included in the model. Table 3: Variables used in explanatory model: Variable Variable Name Source Variable Measure Time Economic Indicators Gross Domestic Product GDP BEA Dependent Thousands, By 1997-2009 State Gross Domestic Product lgdp BEA Dependent By State 1997-2009 Logged Unemployment Unemployment BLS Independent Percentage, By State 1997-2009 GDP per capita GDP_cap BLS Independent Percentage, 1997-2009 By State Education % of people with Bach NSF Independent Percentage, 2007 bachelor s degrees By State Education - % of SE sedegree NSF Independent Percentage, 2007 degrees conferred By State Productivity Academic patents Patents_se NSF Independent Per 1,000 SE 2007 docs, By State Private patents Patents_aca NSF Independent Per 1,000 SE 2007 docs, By State Firm Level Investment Percentage of high-tech Httotal NSF Independent Percentage, 2007 firms in a state By State R&D Tax Credit Presence of Credit Credit Legislation Independent Y/N 1997-2009 Credit Rate taxrate Legislation Independent Percentage 1997-2009 Economic indicators were a key unit of analysis, variables of interest include: Gross Domestic Product (GDP), unemployment, and GDP per capita. My primary dependent economic indicator is GDP by state. The natural log of GDP was computed to provide a comparative analysis of changes and growth between states. Logging GDP allows for a comparison between states regardless of the initial size of GDP. It focuses the analysis on 16

the percentage change, instead of absolute dollar change, which makes it more relevant for this analysis. Unemployment was added as an independent variable because of the potentially significant changes in unemployment that occurred between the years 2007-2009 during the recession. Unemployment is a key indicator of economic growth and prosperity and has a potentially large impact on GDP. GDP per capita was added to the analysis to control for population factors in the analysis. Several types of variables to measure R&D effectiveness were examined, falling into several major categories: education, R&D productivity, and R&D investments. The key education variable captures the overall education level of individuals in the states by measuring the percentage of individuals with bachelor s degrees. Various measures of productivity, in this case the number of patents issues per one thousand science and engineering PhDs, were used to examine how effective and efficient scientists are in their ability to produce new innovations and patent new technologies. Both measures for private patents and academic patents were included in the analysis. Lastly, firm level investments into R&D were measured to account for the varying business environments within each state. The total percentage of firms considered high-tech was measured. Traditionally, these firms were seen as commercial producers, being able to take R&D and turning it into a commercial product. Based on analysis from others, I do not include business tax rate by state. This is because previous empirical analyses (e.g., Wu 2005) found no significant findings regarding the impact of the overall tax rate. Furthermore, it would be challenge to calculate the actual corporate tax rate faced by firms as each firm does not face the same corporate 17

tax rate based on its specific business model. Attempting to control for this within a state would not produce an accurate measurement and therefore, is excluded from the analysis. 18

Chapter 3: Descriptive Statistics Table 4 outlines the descriptive statistics on both the dependent and the independent variables in the model. As described previously, the data are categorized into economic indicators, measures of education, and measures of productivity, private firm makeup, and R&D tax credit information. Of particular note are the wide variations in economic indicators, such as GDP ($14.5 billion to $1.92 trillion) and unemployment (2.3-13.6 percent). The maximum for GDP per capita is Washington, DC; this is an outlier within the data. Table 4: Descriptive Statistics Variable Variable Name# of Obs Mean Std. Devi Min Max Economic Indicators Gross Domestic Product (millions) GDP 663 223471 27411 14553 1921493 Work Unemployment Rate Unemployment 663 5.01 1.60 2.3 13.6 GDP Per Capita GDP_cap 663 41296 15150 25200 148813 Education % of people with bachelor degrees Bach 663 29.42 6.679 19.4 56.4 Productivity Academic patents per 1000 SE docs Patents_se 624 11.46 8.520 1.1 49.8 Private patents per 1000 in SE Patents_aca 663 8.49 4.835 0 24.7 industry Firms % of firms that are high tech Httotal 663 7.94 2.020 4.86 14.6 R&D Tax Credit Information Presence of Credit Credit 663.54.498 0 1 Credit Rate taxrate 663.04.0488 0.2 While basic descriptive statics are important indicators for the data, it is important to compare two scenarios, states with R&D tax credits and states without R&D tax credits. Table 5 examines the differences within the descriptive statistics between states with R&D tax credits and states without R&D tax credits. What the table suggests is that states with 19

tax credits have a higher GDP, lower unemployment, lower per capita GDP, higher education levels, develop more patents, and have a higher percentage of high-technology firms within the states. Table 5: Descriptive Statistic Comparison for States with and without Tax Credits Variable # of Obs w/ Credit # of Obs w/o Credit Mean With Mean Without Difference Economic Indicators Gross Domestic Product (millions) 363 300 252898 187863 65035 Work Unemployment Rate 363 300 4.969 5.062-0.093 GDP Per capita 363 300 39941 42936 2995 Education % of people with bachelor degrees 363 300 30.278 28.385 1.893 Productivity Academic patents per 1000 SE docs 363 300 13.347 7.757 5.59 Private patents per 1000 in SE industry 363 300 9.1 9.024 0.076 Firms % of firms that are high tech 363 300 7.858 2.206 5.652 Figure 1 demonstrates the differences between states with and without a tax credit over time. Per capita GDP is used to demonstrate the difference in GDP while controlling for population. Additionally, Washington, DC is excluded from the per capita GDP analysis because it is an outlier that skews the analysis. 5 As noted in Figure 1, average per capita GDP tends to be higher in states with a R&D tax credit. Yet, in the years during the recession, it appears that GDP per capita drops more in states with a R&D tax credit than states without a R&D tax credit. 5 Annual per capita GDP from the survey was approximately $140k; average per capita in other states was approximately $45k. Appendix 2 presents the per capita difference when Washington, DC is included in the analysis. 20

Figure 1: Comparison of per capita GDP in states with and without R&D Tax Credits. Changes in economic indicators were present when comparing states with credits and without credits. Table 6 classifies states into two categories, those with R&D tax credits and those without R&D tax credits, then compares the changes in economic indicators from just before to just after the recession. During the recession states with R&D tax credits saw GDP increase 6.1 percent v. 5.6 percent for states without tax credits. In contrast, unemployment in states with credits rose 48.5 percent v. 46.5 percent for states without credits. In states with credits, per capita GDP decreased by 2.6 percent, while it was fairly stagnant in states without R&D tax credits (a change of 0.2 percent). 21

Table 6: Dependent Variable Descriptive Statistic Comparison for States with and without Credits in 2006 and in 2009 Variable With Credits Without Credits 2006 2009 Diff % Change 2006 2009 Diff % Change Gross Domestic Product 280229 298403 18174 6.09% 228609 242205 13595 5.61% Unemployment Rate 4.29 8.32 4.04 48.48% 4.64 8.68 4.04 46.51% GDP per Capita 42047 40888-1158 -2.75% 46992 47068 76.16% The descriptive statistics indicate that there was some difference between states with R&D tax credits and states without credits. The results section further explores these potential differences. 22

Chapter 4: Results Regression Findings The multivariate regression was employed to account for key factors in innovation, following the functional form of the other models build by Wu, Wilson, and Hall. 6 I developed one basic model for the analysis (see Model 1) with the dependent variable of GDP and the independent variables of economic indicators, education, patent productivity, private investment, and tax credit information. Models 2, 3, and 4 were subsequently adjusted to account for the years leading up to the recession (1997-2006 for Model 2) and the years during the recession (2007-2009 for Model 4). Because of the greater time period before the recession, Model 3 was constructed to examine the years just before the recession, 2004-2006. The purpose was to capture a time period just before the recession, three years, and the recession time period, three years. The comparisons between Models 3 and 4 provide the basis for my analysis. The multivariate regression model uses panel data on a state-year basis. Panel data allow me to examine the potential impact of polices on a state-year basis. This allows me to account for significant economic changes during the time period examined (Stock and Watson 2006). The model for the analysis is as follows: lgdp = ß 0 + ß 1 unemployment + ß 2 GDP_cap + ß 3 bach + ß 4 patents_se + ß 5 patents_aca + ß 6 httotal + ß 7 credit + ß 8 rate 6 As stated earlier in the paper, key elements of a successful innovation policy include examination of educational attainment, productivity (in this case patents by both academia and the private sector), private investments in technology, and other economic indicators (such as unemployment and GDP per capita). 23

The model includes economic variables, education variables, productivity variables, private market variables, and variables regarding the presence of a credit. GDP, the dependent variable, was logged for the purpose of my analysis. 7 GDP was logged for the following reasons: 1. The distribution of GDP across the states does not appear to be standard (see Figure 2 - GDP distribution). With outliers such as California and the District of Columbia, the boundary wherein GDP lies was great (Wooldridge 2009). 2. To examine the impact of policies in a percentage format. When the analysis was completed in linear formation, the results were more applicable to policy decisions when it can be interpreted as a percent change in GDP versus an absolute. For example, a $50 million increase in GDP means more to policymakers in Alaska than California. However a 5 percent change means the same. Logging GDP fits within the convention of cross-geography examination of GDP and fits the commonly held requirements on when to log a variable (Wooldridge 2009). I did not see the same variation or need to log other variables as they were either reported in a percentage form, or as a calculation (in the case of GDP per capita). 7 To examine data when GDP is not logged, see the appendix. The significance of the other variables in the model did not change when logging GDP. 24

Figure 2: GDP Distribution The results of the explanatory model are presented in Table 7. The results are presented in terms of the natural log of GDP. See Appendix 1 for further test models run, including models without GDP logged, and the use of other variables not used in the final model. 25

Table 7: Explanatory Models with lgdp as the Dependent Variable Explanatory Variable Model 1 Model 2 Model 3 Model 4 All Years Before During Before Recession Recession Recession (2004 2006) (1997-2006) (2007 2009) Economic Indicators Unemployment.1678***.2535***.3005***.0896*** GDP per Capita Education % of population with Bachelor s Degrees R&D Productivity Private patents per 1000 in SE Industry Academic patents per 1000 in SE Docs Private Investments % of firms that are high-tech R&D Tax Credit (.0191) (.0313) (.05598) (.0263) -.000015*** -.00002*** -.000022*** -.00002* (.000003) (.000004) (-.000007) (.000006).0026.0100.0112.0042 (.0067) (.0078) (.0133) (.0137) -.0162*** -.0160*** -.01335* -.0185*** (.0037) (.0043) (.0075) (.0077).1238***.1240***.1187***.1175** (.0065) (.0074) (.0129) (.0397).1606***.1684***.1962***.1744*** (.0200) (.0231) (.0399) (.0397) Presence of a credit.5355***.5300***.3603*.4596*** (.1043) (.1196) (.2048) (.2098) Tax credit rate -3.545*** -3.502*** -2.343-3.869** (1.026) (1.191) (1.913) (1.944) _cons 9.205 8.719 8.551 9.786 R 2 (Adjusted R 2 ).5195 (.5133).5268 (.5188).5750 (.5498).5321 (.5044) Number of Observations 624 480 144 144 *** - Significant at.001 ** - Significant at.05 level * - Significant at the.10 level In my overall model, analysis suggests that the mere presence of a tax credit increases GDP, regardless of the economic conditions (i.e., if there is a current recession or growth in the economy). The analysis suggests that holding unemployment, GDP per capita, education, patents for academia, patents for the private sector, the percentage of firms that are high-tech, and the R&D credit rate constant, results in states with a R&D tax credit see an increase of 0.45 percent in GDP over states without a tax credit ( significant at 26

.001 level). While the findings are significant, it is a slight decrease from the ten years before the recession, wherein holding all other factors constant, an increase in GDP for states with a R&D tax credit was 0.53 percent, a 12 percent reduction in the impact of the R&D tax credit. When examining the three years just before the recession, while holding education, unemployment, GDP per capita, education, patents for academia, patents for private sector, the percentage of firms that are high-tech, and the R&D tax credit rate, the increase in GDP associated with a tax credit is 0.36 percent, compared to a 0.45 percent increase in GDP during the three years during the recession. The results suggest that in an economic recession, the presence of a R&D tax credit does dampen an economic downturn by continuing to increase GDP over states without a R&D tax credit. However, the effect of the credit during a recession is not as high as it is in a period of economic growth, the ten years leading up to the recession, which included the dot-com boom. This is likely due to many factors, including the availability of capital, uncertainty about the economic future of the country and a willingness for firms to take risks in investments, and the debate of the federal R&D tax credit. 8 When isolating the time period for the analysis to the three years just before and the three years during the recession, the results are similar to those shown over all years. Of note, when the time period approaches the recession, the associated increase in GDP due to a tax credit appears to decrease. The ten years before the recession, holding other factors in the model constant, show, on average, a 0.53 percent increase in GDP compared to an 8 During the time of the analysis, the R&D tax credit was heavily debated in Washington, DC. At the center of the debate was not the effectiveness of the credit, but whether the credit should be made permanent so that firms may have more predictability to make decisions in the long run. The R&D tax credit was renewed and in place during the recession, but it wasn t until the spring of 2011 that the administration and Congress sent strong signals about making the tax credit permanent. 27

increase of 0.36 percent with the years leading up to the recession. This is potentially due to a number of factors, ranging from national economic policy, the size of the deficit, and the ending of the dot-com boom. During the recession, the increase in GDP associated, holding other factors in the model constant, with a R&D tax credit is 0.46 percent compared to a 0.36 percent increase with the three years just before the recession. The 22 percent increase in GDP associated with the tax credit is indicative the potential important role the R&D tax credit plays through an economic downturn. The 22 percent increase suggests that during a recession, the R&D tax credit dampens, if only slightly, an economic downturn. The growth seen from implementing a tax credit is balanced with the decrease in GDP associated with increasing credit rates. This suggests a delicate balance when implementing a R&D tax credit. The overall model suggests that for every percentage point increase in a R&D tax credit rate, holding all other factors in the model constant, GDP decreases by 0.035 percent (significant at.001 levels). Analysis of the other models shows similar results. The models examined suggest that the presence of an R&D tax credit increases GDP, yet with every percentage point increase in the credit rate, GDP decreases. The models are consistent in this trend regardless of the economic conditions and the presence of a recession. What changes, however, is the magnitude of the impact of the R&D tax credit. 28

Thoughts for Future Research In my analysis, some of the data were lagging indicators and information was only available for one year, 2007, at the start of the recession. Data for education, patents, and the number of high-technology firms was only available for 2007. For future analysis, I would suggest finding comprehensive data that can track these variables over time. Secondly, many of the variables taken from the National Science Foundation survey are lagging indicators. The number of patents issued in 2007 do not necessarily reflect the work completed in 2007, rather they reflect the work completed in advance of 2007. Therefore, if the recession had an impact on patent productivity, then it may not show up in the data until a survey taken in 2009. The analysis is also limited in that it cannot see the impact of the R&D tax credit after economists declared the recession over (in 2009). A complete comparative analysis of the model in the three years after 2009 should be completed to see how the presence of the R&D tax credit affected GDP after the recession. Did the recession have lagging impact on the amount of R&D investment? What happened to patent productivity? Did the rate of growth in GDP slow as compared with the time period right before the recession? These questions could be answered with analysis of additional data from 2009-2012. 29

Chapter 5: Conclusion Policy Implications Analysis suggests that the state-based R&D tax credit has a positive impact in economic growth from a variety of analysis. However, when examining the results of my analysis, it is important to remember that the states were not randomly assigned to a treatment group; that is, they were not randomly assigned to have a R&D tax credit. Therefore, it is challenging to isolate the effect of the R&D tax credit because many factors went into policymakers decisions to implement the credit. The results suggest that state investment into an R&D tax credit will continue to spur GDP growth over states which do not have an R&D tax credit. As Wilson found in his analysis, states with credit demonstrate higher economic growth than surrounding states, but that growth comes at the expense of other states (2006). My results do not refute this. As my results suggest, an R&D tax credit increases GDP, even in an economic recession. Therefore, the tax credit appears to be a policy option that policymakers should consider when developing economic policy for their state. However, before policymakers begin implementing R&D tax credits, they should consider other important factors, such as: 1 What is the purpose of the R&D tax credit? To spawn immediate growth? 2 What are the types of industries in the state that policymakers hope to have benefit from such a policy? 3 Are policymakers trying to attract new industries and firms to the state? 30

4 What investments are policymakers willing to make in the educational system to ensure that firms have employees who are able to be productive contributors? A R&D tax credit is a signal to firms that a state is serious about making investments in innovation. If a state wishes to implement an R&D tax credit, it should consider making it permanent so that firms can eliminate some of the uncertainty when making investment decisions. The R&D tax credit is one step towards building an economic ecosystem geared towards innovation; however it cannot be done in isolation. Merely implementing an R&D tax credit without considering the other factors that make the credit successful is not the best use of policymakers resources. Conclusion The R&D tax credit will likely be a tool that many states continue to employ to attract innovative firms into their state. The results suggest that it is not only an effective economic policy tool in times of economic growth, but also in times of economic contraction (though with a lesser impact). Policymakers should consider implementing credits as a way to spur economic growth in their states. However, the credit itself should not be seen as a magic bullet that can work in isolation. There are many factors that influence economic policy and a R&D tax credit is just one such tool used to impact the economic environment within a state. 31

Appendix A: Alternative Models Table 8: Estimated Coefficients for GDP in a State Model 1 Dependent Variable: GDP Models 2-6 Dependent Variable: Log of GDP Explanatory Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 _Cons -389277.7 19.874 10.377 10.042 0.959 10.227 Economic Indicators Unemployment 31682***.0707***.1419***.1453***.1009***.1423*** (4957.02) (0.0174) (.0183) (.0182) (.0175) (.0181) % of R&D per GDP -52720*** -.0510*** -.0244 -.0580** -.0856*** (6803.99) (.0232) (.0250) (.0257) (.0240) % of private R&D per GDP.0954*** (.0355) % of Bachelors in Workforce Education % of population with High School Degrees % of population with Bachelor s Degrees % of degrees which are Science and Engineering R&D Productivity Private patents per 1000 in SE Industry Academic patents per 1000 in SE Docs Private Investments % of total firms that are hitech Venture capital expenditure per $1000 GDP R&D Tax Credit 30735***.0614***.0156**.1736***.0037 (4986.72) (.0065) (.0077) (.0176) (.0074) -.1274*** (.0087) -23749.06*** -.0104** -.1277*** (3685.77) (.0057) (.0130) -650.366 -.0360*** -.0400*** -.0411*** -.0368*** (1520.02) (.0057) (.0058) (.0054) (.0058) 442.33 -.0093*** -.0111*** -.0104*** -.0135*** -.0145*** (991.19) (.0034) (.0038) (.0037) (.0035) (.0039) 30156.88***.1141***.1276***.1277***.1333***.1232*** (1838.80) (.0064) (.0070) (.0070) (.0065) (.0070) 5375.60.1318***.1024***.0710***.0966*** (5025.92) (.0179) -0.0188 (.0177) (.0188) 54933.15***.0070.0755***.0712***.0663***.0125 (7156.32) (.0244) (.0271) (.0272) (.0253) (.0291) Presence of a credit 56070.68*.8734***.4521***.4132***.8180***.3220*** (29993.01) (.0964) (.1067) (.1049) (.1059) (.1078) Tax credit rate -313179.60-5.830*** -3.223*** -3.170*** -6.901*** -2.316** (298986.3) (.9693) (1.062) (1.059) (1.056) (1.075) R 2 (Adjusted R 2 ).546 (.538).618 (613).535 (.528).536 (.528).599 (.592).537 (.530) # of Observations 624 624 624 624 624 624 *** - Significant at.001 level ** - Significant at.05 level * - Significant at the.10 level 32