Legislated Incentives and Expectations by Thomas Doran A dissertation submitted to the Graduate Faculty of Auburn University in partial fulfillment of the requirements for the Degree of Doctor of Philosophy Auburn, Alabama May 4, 2013 Keywords: Tariffs, Logistics, Tax Credits, Haiti, Automotive, Veterans Copyright 2013 by Thomas Doran Approved by James Novak, Chair, Professor of Agricultural Economics and Rural Sociology Diane Hite, Professor of Agricultural Economics and Rural Sociology C. Robert Taylor, Professor of Agricultural Economics and Rural Sociology Norbert Wilson, Professor of Agricultural Economics and Rural Sociology Peng Zeng, Professor of Statistics ii Abstract The United States Congress passes several pieces of legislation with the intent to improve the overall livelihood of Americans and assist other nations toward economic development. This collection of essays identifies three acts of Congress which attempt to i) enhance livelihood of Americans through free trade, ii) help the economic development of a poor nation, and iii) attempt to enhance employment opportunities of Americans who historically have low employment rates. Chapter 1 considers the Korean-U.S. Free Trade Agreement (KORUS). The KORUS Free Trade Agreement was created with the intent to expand and secure each country?s global competiveness. The U.S. import market for automotive wire harnesses is examined to evaluate the expected benefits to U.S. importers and South Korean exporters. The results suggest relatively small direct benefits from the removal of tariffs on wire harnesses, which may indicate similar results for other automotive component trade. However, indirect effects from the free trade agreement may give each side of the market greater benefits through flexibility to optimize cost structures. Chapter 2 investigates U.S. efforts to develop the economy of Haiti through removal of tariffs on Haitian produced apparel. The analysis suggests tariff removal for cotton based apparel should be expected to be different from tariff removal for wool and synthetic based apparel. Results suggest potential increases in exports to the U.S., but with limited new opportunities for Haiti?s existing production. A potential for negative impacts for Haitian iii farmers is identified from separate legislation with the intent to enhance the Haitian apparel industry. In general, Haiti?s own development of infrastructure may provide increased income for Haitians through internal transportation cost reductions and provide greater opportunities to more value-added apparel products. The third chapter uses categorical analysis to measure impacts of the Work Opportunity Tax Credit (WOTC) employer hiring incentive program for employment conditions of U.S. veterans compared to their non-veteran counterparts. The results indicate disabled veterans may have gained versus their disabled non-veteran counterparts. Future benefits to disabled non-veterans may come from employment accommodations provided to disabled veterans through the WOTC. Employee reactions appear more influential to program success than reactions of employers. iv Acknowledgments ?In my country, if you could get a job, you?d have to work every day and spend nothing you?ve earned for five years to buy a car, an old [bad] one. No food, no clothes, no cigarettes? I came to this country, worked at this place six months and was able to get my new car, and it?s a good one. In this country you can have anything you want, all you have to do is work for it.? - Srdjan Kliska, 3:00am in a factory parking lot, New Baltimore, MI (1999) I would like to offer sincere appreciation and gratitude to my advisor Dr. James Novak for providing support during my time at Auburn. I am grateful to committee members Dr. Robert Taylor and Dr. Peng Zeng for important comments while working on this project. I also express sincere gratitude for guidance from committee members Dr. Norbert Wilson and Dr. Diane Hite not only this work, but several other projects while I have been at Auburn. In particular, I express the utmost gratitude for Dr. Wilson?s patience and genuine attempts to convert my untrained thoughts to those which could be understood by the academic community. I would also like to thank Drs. Diane Hite, Valentina Hartarska and Nedret Billor for teaching important data-handling techniques which provided exposure to an entire industry I did not really understand even existed prior to coming to Auburn. I offer special appreciation to Dr. Hartarska for giving blunt feedback on a few projects that I will continue to keep in mind while working on others. I want to express sincere appreciation to Dr. Henry Thompson for taking the time to read my work and give constructive criticism, even though I never had a class with him. I want to thank Dr. Richard Beil for his time and patience extracting knowledge from me on a v few important occasions. Additionally, I would like to thank Dr. Joseph Molnar and Dr. Curtis Jolly for simply taking time to make sure I was aware of some events around campus that may interest me and help in my studies while at Auburn. Outside Auburn, I would like to thank Mr. Mike Glynn and Mr. Todd Torres who provided me opportunities to work flexible schedules which were necessary to attend class. Working on my doctorate took a coordination of several personal and professional situations, which could not have been done without Mr. Glynn and Mr. Torres? flexibility and understanding. I also want to acknowledge a few simple lessons in economics and management I learned long before coming to Auburn which stuck with me, and are important in life. My undergraduate professor at Hope College Dr. Robert Gentenaar explained the concept of TANSTAAFL (There ain?t no such thing as a free lunch), which is probably taught the first day of any introductory economics class, but is applicable to everyday life. When I was 13 I had a rookie card of some All-Star future Hall of Fame baseball player. I remember turning to my dad and asking him what he thought it was worth. His response was, ?Whatever someone?s willing to give you for it.? This was probably my first understanding of the difference between price and value. Dr. Anthony Muiderman, also of Hope College, taught the important practice of ?servant leadership? in organizational management. The more I think of it, the more I find the most effective people I?ve encountered practice servant leadership. I also would like to take time to express genuine appreciation for the entire community of Auburn University and the idea of the land grant university. My time at Auburn has been full of great opportunities to attend several seminars on enlightening topics. Every week someone is speaking or presenting material that is open to the public. I have had the opportunity to meet vi foreign dignitaries, civil rights leaders and other scholars most people simply do not have time or opportunities with which to interact. Finally, I have been asked several times why I stopped what I was doing in the private sector to come back to school. I think of a few people and experiences I have come across in the past several years. My high school football coach Gary Steinhoff just mentioned to me one day during practice he thought I could play in college. I had always thought that was something out of my reach. If he would not have said that to me, I would not have even tried and have no regrets regarding it as a result. Srdjan Kliska, with whom I worked early in my career in Michigan, is a Serbian war refugee. He and his father were forced out of his house at gun point by the Serbian army when he was a young teenager, without opportunity to collect any belongings and contact his mother. He eventually managed to make his way to America and reunited with his mother. He is one of several refugees I met from the former Yugoslavia while in Michigan. Working and interacting with them gave me greater understanding of the vast fortunes and opportunities we have in the United States, along with some of our discrimination practices. I met Peter FaFord on my first trip to a small village and orphanage in Haiti. Peter is a Deacon in the Catholic Church, but also had a long career as a skilled project manager and civil engineer. Peter and his wife Linda could live very comfortably in the United States, but have chosen to dedicate a significant part of their lives in Haiti helping those who can offer nothing of tangible value in return. One of Peter?s tasks was to remove standing water by laying a pipe with a ?? drop for every one foot in length for several hundred feet until the pipe reached a small body of water, away from the orphanage. The task was considerable given the terrain, but also given he had to design the project keeping in mind scores of children between the ages of 2 and 6 years-old who love to run around and jump on things, including PVC pipe. The pipeline vii removed many mosquitoes and disease-carrying bacteria from hundreds of people. It directly and efficiently reduced exposure of several contagious diseases. The resources which could be allocated toward education and other efforts toward developing the country, rather than correction of a disease condition, would be difficult to measure for any economist. Peter?s experience and knowledge became very practical in a situation long after he gained them. I have come across several stories of efforts to assist Haitian development over the past few years, most of which were high level and abstract with varying degrees of success or failure, let alone feasibility. I cannot think of one with as much effectiveness as what Deacon FaFord offered. I met several people in my time working in the Detroit area who had years of experience and for whatever reason did not further develop their skills and technical knowhow to survive in the ever-changing industry. I can recall several times when someone?s security badge suddenly would not work, which became known as notice of termination. Those who actively sought to improve their skill sets seemed to survive. I am not exactly sure why I decided to pursue a doctorate, but I believed I could, I had the opportunity to do it, I figured it would be practical to a situation somewhere down the road, I believe it is important to be ahead of the curve of practical work skills, and I did not want to be 80 years-old and have regrets about not even trying. Somewhere within all that, I decided take a few years on work on my doctorate. War Eagle! viii Table of Contents Abstract ......................................................................................................................................... ii Acknowledgments........................................................................................................................ iv List of Tables ............................................................................................................................... xi List of Figures ............................................................................................................................ xiv List of Abbreviations .................................................................................................................. xv Chapter 1 Free Trade Agreements in the U.S. Automotive Subcomponent Import Market ......................... 1 1.1 Introduction ............................................................................................................... 1 1.2 The Automotive Industry and Wire Harnesses ........................................................... 3 1.2.1 The Automotive Industry ............................................................................. 3 1.2.2 Wire Harnesses ............................................................................................ 4 1.3 Model Specification Estimate ..................................................................................... 5 1.4 Data for Estimation ..................................................................................................... 7 1.5 Elasticity Estimations???????????????..???????? ....10 1.6 Expected Changes in Korean Wire Harness Price and Imports..????????12 1.7 Summary and Comments???????????????????????1 5 ix Chapter 2 Logistics Performance and Tariffs in the U.S. Apparel Import Market ..................................... 28 2.1 Introduction???????????????????????????? 28 2.2.1 Haiti???????????????????????????. 28 2.2. Literature Review????????????????????????? . 29 2.3 OLS to Selection Model Progression?????????????????.. .. 32 2.3.1 Zero-Inflated Count Model?????. ??????? ??????. 33 2.3.2 Heckman Selection????????????????????. ... 35 2.4 Data for Estimation????????????????????????.. . 36 2.5 General Apparel Imports??????????????????????? 37 2.6 Cotton Apparel Imports ............................................................................................ 41 2.7 Concluding Analysis????????????????????????..43 2.8 Summary and Implications of Findings?????????????????.45 Chapter 3 The Work Opportunity Tax Credit ............................................................................................. 61 3.1 Introduction ............................................................................................................... 61 3.2 Legislative History .................................................................................................... 64 3.3 Theoretical Framework ............................................................................................. 66 3.4 Data and Empirical Analysis .................................................................................... 69 3.5 Regression Estimates ................................................................................................ 71 3.5.1 Regression Results ..................................................................................... 74 3.6 Regression Estimates for Trade Industry Sector ....................................................... 75 3.6.1 Regression Analysis?????????????????????76 x 3.7 Conclusions and Analysis ......................................................................................... 78 Bibliography .............................................................................................................................. 96 Appendix A?. ............................................................................................................................ 99 Appendix B???? ????????????????????????????102 Appendix C????????????????????????????????104 xi List of Tables Table 1.1 Custom Value Market Share and Summary Statistics ????????????1 7 Table 1.2 Custom Value and Dutiable Percentage ............................................................ ?.?1 8 Table 1.3 Model Restrictions ...................................................................................................... 19 Table 1.4 Model Coefficient Results?????????????????????? ?20 Table 1.5 Compensated Elasticities ............................................................................................ 21 Table 1.6 Uncompensated Elasticities ........................................................................................ 22 Table 1.7 Expenditure Elasticities .............................................................................................. 23 Table 1.8 Tariff Removal Incidence Estimates????????????????? ??24 Table 1.9 Year Ending Q3 Wire Harness Exports to the United States?????????..24 Table 2.1 USA Import Market for HTS Chapter 61 Apparel ..................................................... 48 Table 2.2 Descriptive Statistics for General Apparel Imports ????????????? 49 Table 2.3 Variable Correlation Matrix Full Data Set ................................................................. 50 Table 2.4 General Apparel Regression Results .......................................................................... 51 Table 2.5 General Apparel Repeated Heckman Results ............................................................. 52 Table 2.6 Descriptive Statistics for Cotton Apparel Imports ...................................................... 53 Table 2.7 Cotton Variable Correlation Matrix Full Data Set...................................................... 54 Table 2.8 Cotton Apparel Regression Results ............................................................................ 55 Table 2.9 Cotton Apparel Repeated Heckman Results ............................................................... 56 Table 2.10 Animal Apparel Repeated Heckman Results ............................................................ 57 xii Table 2.11 Cotton Apparel Market by Source Country 2007-2011??????????? 58 Table 2.12 Marginal Effect of Perc in Cotton Apparel Market????????????...58 Table 2.13 Export by Apparel Type???????????????????????58 Table 2.14 Cotton Coats, Sweaters and T-Shirts Marginal Effects 2007-2011??????..59 Table 3.1 Work Opportunity Tax Credit Summary . ????????????????. ..85 Table 3.2 Demographic Variable Summary ............................................................................... 86 Table 3.3 Educational Summary ................................................................................................. 87 Table 3.4A Disabled Veteran Employment and Earnings .......................................................... 88 Table 3.4B Disabled Employment and Earnings ........................................................................ 89 Table 3.4C Non-Disabled Employment and Earnings ................................................................ 90 Table 3.4D Veteran Employment and Earnings ......................................................................... 91 Table 3.5 Model Fit Criteria ....................................................................................................... 92 Table 3.6 Wage and Employment Differences Summary .......................................................... 93 Table 3.7 Trade Industry Wage Distributions Fit ....................................................................... 94 Table 3.8 Trade Industry Wage and Employment Differences Summary .................................. 95 Table 3.9 Disabled Hourly Wages .............................................................................................. 69 Table 3.10 Disabled Weekly Wages ........................................................................................... 70 Table 3.11 Disabled Binary Employment ................................................................................... 71 Table 3.12: Disabled Baseline Category Logit ........................................................................... 72 Table A1.1: Automotive Wire Harness Related HTS Codes???????? ..?????99 Table B2.1: HTS Aggregation Level Example??????????????????. .102 Table B2.2: Endogeneity Tests????????????????????????.102 Table B2.3: Cotton Apparel by Commodity Type?????????????????103 xiii Table C3.1: Hourly Wages.?.?????????????????????...?104 -105 Table C3.2: Weekly Wages????????????????????????106 -107 Table C3.3: Binary Unemployment?????????????????????108 -109 Table C3.4: Employment Category?????????????????????110 -112 Table C3.5: Trade Industry Hourly Wages?????????????????? ?.. ..113 Table C3.6: Trade Industry Hourly Wages????????????????????114 Table C3.7: Binary Unemployment?????????????????????115 -116 Table C3.8: Trade Industry Employment Category???????????????117 -119 Table C3.9: Employment Category Comparisons Between Overall Economy and Trade Panel A: Overall Economy ........................................................................................... 120 Panel B: Trade Industry ................................................................................................ 121 xiv List of Figures Figure 1.6: Incidence Scenarios .................................................................................................. 14 Figure 1.1: Wire Harness Imports 1989-2012?????????????????? .?.25 Figure 1.2: Wire Harness Total Expenditure ..................................................................... ?.? 25 Figure 1.3: Wire Harness Market Share ..................................................................................... 26 Figure 1.4: Wire Harness Quantity ............................................................................................. 26 Figure 1.5: Wire Harness Price ......................................................................................... ??.. 27 Figure 1.7: Change in Welfare ................................................................................................ ?27 Figure 2.1: Homoskedasticity Check Panel A: Level OLS model without observations of zero .............................................. 60 Panel B: Logged OLS model with observations of zero................................................. 60 Figure 3.1: Change in Wage Received and Wage Paid for Disabled Veteran ............................ 77 Figure 3.2: Short-term and Long-term Potential Effects ............................................................ 81 Figure A1.1: Wire Harness Example .......................................................................................... 99 Figure A1.2: Korean Wire Harness Imports ............................................................................. 100 Figure C3.1: Hourly Wage Distributions .................................................................................. 122 Figure C3.2: Weekly Wage Distributions ................................................................................. 122 Figure C3.3: Usual Hours per Week ......................................................................................... 123 Figure C3.4: Employment ......................................................................................................... 123 xv List of Abbreviations ADA08 Americans with Disabilities Act of 2008 ADA90 Americans with Disabilities Act of 1990 AIC Akaike Information Criteria AIDS Almost Ideal Demand System Model ARRA American Recovery and Reinvestment Act of 2009 ATPRA American Tax Payer Relief Act of 2012 CAFTA Central America Free Trade Agreement CBS Central Bureau of Statistics Model CBERA Caribbean Basin Economic Recovery Act of 1983 CEPII Centre d'Etudes Prospectives et d'Informations Internationales Prospective Studies and International Information Center CIF Cost Insurance Freight CPS Current Population Survey CV Custom Value DDD Difference-in-Difference-in-Difference EIAP Earned Import Allowance Program FMVSS Federal Motor Vehicle Safety Standards FOB Freight on Board FTA Free Trade Agreement xvi GAO General Accounting Office / Government Accounting Office GDP Gross Domestic Product HELP Haitian Economic Lift Program HOPE Haitian hemispheric Opportunity through Partnership Encouragement Act HMC Hyundai Motor Company HTS Harmonized Tariff System JIT Just In Time JIS Just In Sequence KORUS Korean-U.S Free Trade Agreement LDP Landed Duty-Paid Value LPI Logistics Performance Index NAFTA North American Free Trade Agreement NBR National Bureau of Research Model OEM Original Equipment Manufacturer OLS Ordinary Least Squares RHS Right Hand Side ROW Rest of World SSDI Social Security Disability Insurance SME Square Meter Equivalent SNAP Supplemental Nutrition Assistance Program USITC United States International Trade Commission VOW11 Vow to Hire Heroes Act of 2011 WOTC Work Opportunity Tax Credit xvii ZINB Zero-Inflated Negative Binomial ZIP Zero-Inflated Poisson 1 Chapter 1 Free Trade Agreements in the U.S. Automotive Subcomponent Import Market 1.1 Introduction The U.S. automotive industry has shown considerable recovery since large volume declines between late 2008 and mid-2009. The recovery has driven increased demands for imported subcomponents, including automotive wire harnesses. Figure 1.1 shows the top 15 sources for wire harnesses imports in terms of quantity between 1989 and 2012. Mexico is by far the largest source with over 80 billion units imported over the time period. As the demand for wire harness imports has increased, U.S. wire harness exports and employment have steadily declined, along with many other automotive subcomponents (Klier & Rubenstein). The U.S. Census Bureau indicates employment for automotive electrical component manufacturing declined as much as 40% between 1997 and 2007 (2011)1. The United States International Trade Commission (USITC) reports an automotive wire harness trade deficit with all partners grew from just under $5 billion to well over $8.7 billion between 2003 and 2012 Q3, a 75% increase (2012). A review of U.S. based wire harness manufacturers listed with the Wire Harness Manufacturers Association showed several manufacturers of wire harnesses in the United States, but mainly for industries other than automotive (2013). The ones with automotive capabilities were mostly geared toward small scale service production. These conditions indicate U.S. automotive sales growth will be supported through the wire harness import market, rather than domestic wire harness production. Figure 1.1 can be segmented into three types of countries. Mexico stands by itself as the dominant market leader, taking advantage of its proximity and free trade status through NAFTA. 1 North American Industry Classification System code 336332 includes wire harnesses. Employment in within this code is reported at 97,572 in 1997 and 58,823 in 2007. 2 Germany, Japan and South Korea are countries with passenger vehicle assembly companies located in the United States. South Korea has the distinction of being the lone country with assembly locations in the United States aided by a new free trade status for automotive components. All other countries can make up the rest of the supplier side of the wire harness market, and are for the most part low wage countries. The United States and South Korea signed a free trade agreement (KORUS) in June of 2007 removing tariffs on 95% of industrial and consumer goods (U.S. - Korea Free Trade Agreement). The agreement took effect on March 15, 2012, with some adjustments since the 2007 agreement. A major segment of the agreement involves the automotive industry. KORUS phases out a tariff of 2.5% on vehicles from South Korea over the five years following the agreement?s ratification, but removes tariffs immediately on several intermediate components entering the U.S. from South Korea for use in the growing South Korean automotive assembly plants in Alabama (Hyundai) and Georgia (Kia). KORUS contains a ?snapback? clause which permits the U.S. to reapply the 2.5% vehicle tariff should a surge of South Korean vehicles enter the U.S. market. Beyond tariff incentives, HMC2 has also seen recent upward pressure on its domestic automotive labor costs (Woodyard 2012). The 2.5% tariff on assembled passenger vehicles with removal of tariffs on automotive components would seemingly encourage greater U.S. production of the HMC vehicles. KORUS had been in negotiation, or at least feasibility analysis, from as early as 1999 (Cheong and Wang 1999). Lee and Lee use highly aggregated data and project a 2.8% increase in South Korean domestic automobiles and components in the short-term, and a 4.1% increase in the long-term (2005). Lee and Lee also project a 17.5% increase in employment in the South 2 Hyundai Motor Manufacturing of Alabama (HMMA) and Kia Motor Manufacturing of Georgia (KMMG) are subsidiaries of Hyundai Motor Company (HMC). 3 Korean auto sector as a result of the then projected FTA. The present study estimates the expected change in imported components from removal of tariffs using disaggregated data for a single component, wire harnesses. 1.2 The Automotive Industry and Wire Harnesses 1.2.1 The Automotive Industry A brief overview of the automotive industry may give insight to future expectations about procurement managers? reactions to price, regardless of the supplier?s location and product. Original equipment manufacturers (OEM) coordinate hundreds of suppliers and thousands of components from across the globe for one vehicle. Suppliers that supply the OEM are considered Tier 1 suppliers. Suppliers that supply the Tier 1 suppliers are considered Tier 2 suppliers, and so on. Production forecasts can typically be 10 to 12 weeks, give or take a couple weeks depending on the OEM and commodity type. The production forecast is typically disseminated down through the supply chain on a weekly basis. As supply comes back up through the supply chain to the OEM, the commodity becomes more particular to the vehicle. This means each supplier along the Tier system may be the only facility in the world with production equipment and tooling built for the OEM?s specific vehicle. Delays of supply to the OEM are simply not tolerable due to extensive labor and inventory costs that cannot be avoided through the supply chain. For instance, a vehicle cannot be sent to the dealer without a steering wheel. If the steering wheel supplier does not supply the steering wheel, the whole process eventually stops. Ultimately, the costs of delaying production at an OEM assembly plant can range up to and beyond $1000 per minute. A sense of urgency is passed down from the Tier 1 to the Tier 2 and so. Any procurement manager considering a new supplier, or new supplier location, must keep this in mind. Also, the OEMs control vehicle design and assembly, so 4 agreement from the OEM to change production process or location is often required, and not always granted. A rule of thumb for automotive OEMs has traditionally been to keep roughly 60 days of finished goods inventory between the assembly plant and the dealership. At the same time, an emphasis on inventory reduction occurs from the OEM receiving dock back through the supplier tier system, or a just-in-time production (JIT) system, where delivery up through the supply chain occurs on weekly, daily or even less than hourly basis. Further, more complex components may require just-in-sequence (JIS). The JIT/JIS production system mandates reliability for delivery time and has very low tolerance for defects (Wagner and Silveira-Camargos 2011). The U.S. auto industry had 1.25 million units sold in August of 2008. That number dropped to less than 750,000 by November before eventually bottoming out in January 2009 at 657,000 units (Automotive News Data Center 2012). This sudden drop in sales resulted in large inventory increases in terms of holding costs, as well as simple physical space. Components from overseas in Europe and Asia were particularly troublesome because of 8 to 12 lead times on deliveries. The eventual recovery of auto sales helped deplete inventory, but also left purchasing and procurement managers with a heightened sensitivity to inventory commitments. 1.2.2 Wire Harnesses A wire harness has a unique role in a vehicle. It is a collection of electrical wires and plugs connecting occupant controls, and other equipment, to the vehicle?s various mechanical functions. The wires are typically bundled together with a flexible tape or tube by hand, making it a very labor intensive product and appropriate for continued outsourcing to foreign countries. This makes the wire harness potentially atypical for analysis in the automotive component sector. 5 Wire harnesses interact with many sophisticated technologies, such as airbags, but they require rather low technology and labor costs to produce. This separates them from other components, such as stampings or molded products which require large capital investments. Wire harnesses require relatively low capital to set up, thus seemingly allowing for ease of production transfer. However, due to the interaction of the wire harness with safety and key functional components, its production location can have some stickiness. Any vehicle must undergo numerous rounds of testing according to Federal Motor Vehicle Safety Standards (FMVSS), along with OEM and industry specific standards. Wire harnesses serve as the ?nerve? system of the vehicle in that they connect the driver controls to the various electrical components around the vehicle, and thus greatly affect vehicle conformance results. Delivery, production and regulatory requirements can be seen as entry barriers beyond competitive pricing for wire harness imports. 1.3 Model Specification Estimate The reaction of wire harness quantity imported from the identified countries from a change in price of the imported wire harnesses is estimated with a form of Barten?s Synthetic model (1993). The model nests four differential demand systems models with the use of two indicator variables. The Synthetic model is often used as an instrument to determine between four commonly used differential demand functional forms: the Rotterdam (Theil, 1965), Central Bureau of Statistics (CBS) model (Keller & van Driel, 1985), the NBR (Neves, 1994) model and the differential form of the Almost Ideal Demand System (AIDS) model (Deaton & Muellbauer, 1980). Matsuda (2005) uses the Synthetic model to estimate Japanese demand for non-durable 6 goods and services. As explained by Matsuda, the Synthetic model itself can be used to estimate the system, not just identify the functional form of the system. As Matsuda explains Barten?s Synthetic model, the expenditure share-weighted change in natural logarithm of qi as represented in Equation 1.1), 1.1) where and . The expenditure share of wire harness is wi=(piqi)/M , where is total expenditure. The term dln is the change in the natural logarithm of quantity qi or price pj. DQ is the Divisia Volume index, , which is intended to represent the real expenditure (or real income in consumer terms) of the wire harness importers. The parameter ?ij is the Kronecker delta where ?ij = 1 if i=j, and ?ij = 0 if i?j. ?i indicates the marginal change in expenditure on wire harnesses and ?ij indicates price effects. The indicator variables ? and ? determine the functional form of the estimation equation. For this paper, intercept and seasonality variables are added for estimation, along with a variable for the growth in U.S. assembly plants from Auto countries. When ?=0, the Synthetic model system treats expenditure effects as a constant, as in the Rotterdam or NBR models. If ? is set to 1, the system allows marginal expenditure effect ?i to vary with expenditure share wi, as with the CBS or AIDS models. If ?=0, the Synthetic system treats marginal substitution effects from price pj as a constant, as with the Rotterdam or CBS models. If ?=1, price effects vary with expenditure shares wi and wj same as with the AIDS or NBR models. A summary of the nested models? functional forms is below: Specified Model Expenditure Effect ?i Substitution Effect ?ij Rotterdam (?=0, ?=0) Constant Constant CBS (?=1, ?=0) Varies with Budget Share Constant NBR (?=0, ?=1) Constant Varies with cross-Budget Share AIDS (?=1, ?=1) Varies with Budget Share Varies with cross-Budget Share 7 1.4 Data for Estimation Data for estimation come from the USITC for each quarter from 1989 through the third quarter of 2012. Three variations of wire harnesses are aggregated from each country for use as the dependent variable. The USITC reports imports of wire harnesses according to Harmonized Tariff System (HTS) codes3. The reported custom value (CV) of each code serves as both the value of the wire harness and the quantity in the USITC report. In other words, wire harnesses are imported according to value, not weight, pieces or another quantitative measure. The landed duty paid-value (LDP) is the sum of customs value plus all other costs associated with bringing the wire harness from the exporting country to the U.S. port of entrance into the United States. This study assumes the law of one price as explained by Lamont & Thaler (2003) for wire harnesses, in that a wire harness imported from Brazil is the same price as one from India. The difference between the two sources is only the difference in the cost of transporting the wire harness from the respective countries, be they from labor- or capital-intensive production processes. Therefore, the price of a wire harness is considered the percentage above custom value made up from delivery costs paid by the importer, 1.2) pit = where pit represents the price of wire harnesses from country i at time t, and j represents each of the three wire harnesses. The custom value is treated as equal to the quantity qijt in the denominator of Equation 1.2) and throughout this study. 3 HTS 8544300000 volume represents over 80% of the volume during the analyzed time period and was increased to a rate of 5.3% for 2013. USITC descriptions are somewhat vague and could apply to product for use in other than passenger vehicles. Figure A1 shows wire harness imports from Korea jump up near 2004 and 2009, in line with the opening of HMMA in 2004 and KMMG in 2008, signaling the proper HTS codes are used for estimation. 8 Three ?countries? are identified for estimation purposes. Mexico (Country 1) is by far the largest exporter of wire harnesses to the United States in terms of custom value with nearly 70% of the U.S. import market over the period from 1989 through 2012. However, its market share has dropped considerably over the past two decades. Figure 1.2 shows expenditure (pit x qit) growth in the U.S. wire harness import market. Total expenditure4 grew 130% from slightly over $3 million to over $7 million between 1992 and 2012. Volume of wire harness imports increased nearly 600% over the same time period, reflecting the large reduction of cost to import a wire harness. Foreign countries with companies assembling passenger vehicles in the United States are identified in the estimates as ?Auto? nations (Country 2). Japan, Germany and South Korea, account for 3.9%, 1.0% and 0.5% of wire harness imports, respectively. The rest of the world (ROW) countries (Country 3) are made up from all other countries exporting wire harnesses to the U.S. since 1989. Table 1.1 displays market share in terms of custom value for the analyzed period, along with summary statistics for the aggregated countries. Table 1.2 displays a breakdown of the three HTS codes imported from the top 15 exporters listed in Table 1.1 only for the year 2011 rather than the overall sample mean values. Note the countries do not change, just the ordering of ranks. The USITC also reports the dutiable amount of the CV for each code. Over half of all wire harness volume is brought into the United States through Mexico under the HTS 8544300000 code, which has the highest tariff, but is mostly brought in without tariff through the NAFTA agreement. Though, well over $1 billion worth of wire harnesses was subject to the 5% tariff. 4 Expenditure refers to the expenditure toward importing wire harnesses, not the expenditure on wire harnesses themselves. 9 Equation 1.1) is estimated using SAS 9.2. Table 1.3 displays specification test results for the nested models, as well as, results for the Synthetic model. The homogeneity restriction is maintained in each of the nested models and the Synthetic model. Symmetry is not strongly rejected in any of the models with the homogeneity restriction imposed. The Synthetic model is also tested for various specifications for an indication of which nested model may most properly indicate the functional form of the estimating system of equations. These results indicate the CBS model specification to most closely match that of the unrestricted Synthetic model. Estimation results of Equation 1.1) with homogeneity and symmetry restrictions are available in Table 1.4. As indicated from pre-estimation tests, the Synthetic model?s ? and ? coefficients are estimated close to 1 and 0, respectively, resulting in similar estimates to the CBS specification. Three additional parameters are estimated for informational purposes. The ?i terms are intercepts, which would indicate a trend in importers? patterns if statistically significant. Estimation results indicate no particular trend of U.S. importers toward or away from any of the countries. The si coefficients are for quarterly seasonality considerations. Each of the model specifications indicates no seasonality effects, as expected due to the nature of the share-weighted dependent variable. The ci parameters are used to indicate the impact of the growth of foreign-owned vehicle assembly plants. The number of assembly plants with headquarters in ?Auto? nations, also known as ?New Domestics?, grew from 6 in 1989 to 16 in 2012. Each model estimates these parameters as not statistically significant. Finally, indicator variables (not reported) were used to control for the large downturn in the U.S. auto market from Q4 2008 through Q2 of 2009. These were estimated as not statistically significant. 10 1.5 Elasticity Estimations Coefficients from Equation 1.1) and expenditure budget shares can be used to find compensated price-elasticities ?ij and expenditure- elasticities ?i (Matsuda, 2005) in Equation 1.3) and Equation 1.4), respectively, 1.3) ?ij = ?ij/wi ? ?(?ij ? wj) 1.4) ?i = ?i/wi + ? Uncompensated price elasticities ?ij can be found through the Slutsky equation, 1.5) ?ij = ?ij - wj?i. Table 1.5 has compensated (Hicksian) own- and cross-price elasticities, calculated from the 1992 Q3, mean, and 2012 Q3 budget shares. The Synthetic and CBS models are considered to represent the functional forms of wire harness imports, based on estimations of the unrestricted nesting coefficients. All the own-price elasticities are in the expected direction. Cross-price elasticities show a tendency toward a complementary relationship for each country. The exception is the only consistently statistically significant substitutable relationship between Mexican and Auto wire harnesses. The impact of model specification becomes apparent when reviewing across Table 1.5. Allowing the synthetic model?s estimate of ? to be the true value suggests the Rotterdam and CBS specifications to be closest to the true model for price expectations. In comparison, the NBR and AIDS models would tend to overestimate the impact of prices on quantity. For the most part, the overestimation has minor impacts on expectations, with the above noted exception between Auto and Mexican wire harnesses, ?12 and ?21. The NBR and AIDS models suggest Mexican and Auto wire harnesses go from reasonably strong substitutes in 1992 to complements 11 in 2012. However, the models holding price effects constant, as in the Synthetic model, suggest the relationship remains at one of substitution. Table 1.6 shows the long-term uncompensated (Marshallian) elasticities are more responsive to price changes, as expected compared to the short-term Hicksian elasticities. Most of the Marshallian elasticities between the three countries are statistically significant at the 5% level. Own-price elasticities are in the expected direction. The cross-price elasticities suggest wire harness imports are clearly complementary in nature. Own-price elasticities for Auto nations indicate minimal changes in quantity from changes in price. This result is consistent across model specifications and time. Table 1.6 also indicates the complementary relationships change over time. For instance, the 1992 Synthetic model results show a 1% increase in the price of Auto wire harnesses decreased the amount of Mexican wire harnesses by 0.1%. Twenty years later, the estimate changes to an estimated decrease of only 0.04%. A similar change can be seen between Auto wire harnesses prices and ROW quantities. Further, the Synthetic model of 1992 shows Mexico?s own-price has greater influence than the ROW price on Mexican quantity. However, this condition flips in the 2012 model, indicating Mexico?s export quantity is affected more by the ROW price than it is by its own-price. This condition can be found across each nested model to varying degrees depending on whether price effects are allowed to vary with expenditure share. In general, this indicates Mexico?s exports to the United States are growing more dependent on ROW pricing than it is on its own or that of Auto countries. This change can be seen with review of Figures 1.4 and 1.5. Mexico?s price went from $4.23 to $1.51 per 100 wire harnesses from 1993 to 2012, a 64% decrease. Mexico?s exports grew from nearly 433 million units to almost 1.5 billion over the same time, a 245% increase. ROW price went from $6.10 to $4.73, only a 22% decrease. 12 ROW quantity went from 144 million to almost 900 million in this same time, an increase of over 500%. All this could be a reflection of decreasing costs of doing business overseas. Expenditure elasticities are in Table 1.7 and indicate near unitary relationships between percentage changes in expenditure and imported wire harnesses from ROW and Mexico. However, Auto countries see a relatively smaller increase in imports from increases in expenditure. The Rotterdam and NBR models, which hold expenditure share constant, shows Mexican expenditure elasticity is increasing and ROWs is decreasing, while they are more likely remaining constant over the past two decades. 1.6 Expected Changes in Korean Wire Harness Price and Imports As noted in the introduction, automotive procurement managers cannot move production locations quickly simply due to price. Table 1.4 and Table 1.5 reflect this type of mindset and activity in the auto industry. Importers do not react much to price in the short-run, as reflected in Table 1.4 compensated elasticities. Price is not necessarily as important as reliability when considering a wire harness in a vehicle. Safety, traceability, durability and compatibility with other sophisticated, high-priced components are critical and represent barriers to entry for the automotive industry. After time and transfer of knowledge, managers can react to price, as reflected in Table 1.5, though the decision still seems not influenced greatly by price. Based on information discussed in section 1.1, the U.S. effectively does not have a domestic wire harness supply of significant magnitude, and is not considered for supply. Using the Synthetic model results in Table 1.6, the estimated percentage change in Korean wire harness imports can be found by allowing it to follow the pattern of the Auto companies: 1.6) DK* = ?21PM* + ?22PK* + ?23PR* 13 where DK is the demand for Korean wire harnesses. PI is the respective price for Mexico, Auto and ROW, with South Korean price represented by the Auto country price. Percentage changes of both sides of the market are represented with (*)5. Assuming no change in Mexican or ROW pricing and permitting the average own-price elasticity from Table 1.6 as statistically significant results in: 1.6?) DK* = -.142PK* Table 1.2 shows South Korea exported nearly 119 million wire harnesses in 2011. Well over 87 million of them were already imported under the zero tariff HTS code. As a result, $1.27 million in tariffs represent 1.1% of the custom value, or $1.10 per 100 units. The minimum price for importing from Auto countries is 4.1% according to Table 1.1, or $4.10 per 100 units. Using this price point, removal of the tariff represents a 27% decrease in price, suggesting the average cost to South Korean exporters is $3.00 per 100 units, from other export costs. The mean share demand elasticity from Table 1.6 and some assumed supply elasticities can provide insights to which side of the market will benefit more from KORUS, 1.7) 1.8) where P and PK represent the changes in price from pre-KORUS equilibrium P0 to the South Korean exporter (cost) and the decrease in price to the U.S. importer, respectively, and T is ad valorem tariff. From Equation 1.7) and Equation 1.8) the incidence of the price reduction for U.S. importers and South Korean exporters depends on the U.S. demand side elasticity ?, as well as, the South Korean side supply elasticity ?. If ?=|?|, then the $1.10 per 100 units is split evenly between the supply and demand sides of the market. Otherwise, the less elastic side of the market gains the overall greater share of tariff removal. This is depicted below: 5 Derivation available in Appendix A 14 Figure 1.7: Incidence Scenarios Table 1.8 displays the price wedge estimates of increasing supply elasticity estimates. Considering wage pressure in South Korea, the post-KORUS supply elasticity may be less than the absolute U.S. demand elasticity. Figure 1.7 exaggerates an estimate of South Korean supplier and U.S. importer welfares. The South Korean supplier appears to garner a greater portion of the total welfare, at least on the surface. In total, the surplus increases on either side of the market are relatively small considering the 2011 custom value of imports was nearly $8.5 billion. Table 1.7 suggests South Korean price decreases will increase supply from both Mexico and ROW, adding to U.S. importer surplus from the post-KORUS condition, ceteris paribus. However, South Korean suppliers may gain more from KORUS than what is readily apparent as well. Reverting back to Table 1.2, Auto countries supply more wire harnesses under the 0% tariff code (HTS2040) than the 2.5% tariff code (HTS4040). HTS2020 and HTS2040 have similar descriptions. Speculatively, the difference in wire harnesses may only be the absence of a couple components on the HTS2040 wire harness that could be assembled to produce an HTS4040. If gross benefit to the South Korean supplier were less than 2.5%, the HTS4040 tariff amount, the tariff on HTS4040 discourages the value-added assembly process. After KORUS, the South Korean supplier can accept gross benefits less than 2.5% and export P Q S 0 S 1 P 0 P P K D If the n pr ice d ec rea se to U.S. importer , P K , is gr e a te r tha n inc re ase of pric e re ce iv ed b y Kor ea n e xpor te r, P P Q S 0 S 1 P 0 P P K D If = the n pr ice d ec rea se to U.S. importer , P K , is equa l to in cre ase of p ri ce recei ved b y Kor ea n expor te r, P 15 under a potentially more preferable HTS code. Theoretically, this could also occur with the wire harness assigned a 5% tariff. These relatively quick changes may be of more immediate benefit to South Korea than the changes in U.S. demand suggested from the compensated elasticities in Table 1.5. And, since the elasticities in Table 1.6 were merely treated as statistically significant for the purposes of welfare analysis, these may be the only realized benefits by the South Korean exporter. Some evidence of this change in production sequence may be apparent in Table 1.9 displaying exports to the U.S. through the first three quarters of 2011 and 2012 per HTS code for the three Auto countries. Japan and South Korean exports grew considerably, while Germany?s decreased considerably. Between the two countries with export growth, South Korea redistributed toward the two HTS codes with tariffs removed after KORUS. At the same time, Japan may have handled growth by shifting toward the lower tariff HTS. This may be an indication of South Korean suppliers gaining further surplus through reorientation of the production process. 1.7 Summary and Comments The Korean-U.S. FTA removes tariffs on several South Korean imported products effective May 2012. A 2.5% U.S. tariff on South Korean passenger vehicles will be phased out in the five years following May 2012. However, KORUS removes tariffs on imported South Korean automotive components immediately, including tariffs on automotive wire harnesses. Three types of countries were used to differentiate U.S. importers? reactions to changes in price, and thus the removal of the wire harness tariff. A Synthetic differential demand system model suggests U.S. wire harness importers have quite inelastic reactions to prices, and thus wire 16 harness import quantities will grow only slightly as a direct result of the tariff removal. Much greater growth of South Korean exports will come due to overall growth of U.S. passenger vehicle production. This finding counters those of pre-agreement estimates for the auto industry, though are limited to only one particular commodity. Welfare estimation using a South Korean supply elasticity less than the absolute value of the U.S. demand elasticity suggest both countries will have surplus gains, with slightly more going to the South Korean supplier, though both gains are rather negligible. However, each country may also have surplus gains indirectly from the KORUS FTA. South Korean suppliers may gain efficiency through a more flexible production system, while U.S. importers should realize benefits from Mexico and rest of world wire harness suppliers. Wire harnesses represent only one of scores of automotive components imported for assembly in the U.S. passenger vehicle supply chain. This study may offer some indication of the impact KORUS has on the U.S. automotive market prior to the removal of tariff on South Korean completely assembled passenger vehicles. Increased exports from South Korea will come due to growth of the vehicles produced and sold in the U.S. by South Korean assembly companies, not due to price reductions from KORUS. 17 Table 1.1: Custom Value Market Share and Summary Statistics Country Custom Value Market Share 1 Mexico 83,302,476,367 67.90% 2 Philippines 6,343,494,957 5.17% 3 China 4,964,077,964 4.05% 4 Japan 4,744,072,878 3.87% 5 Taiwan 4,701,613,537 3.83% 6 Honduras 3,087,093,618 2.52% 7 Thailand 3,016,366,642 2.46% 8 Canada 2,736,330,374 2.23% 9 Nicaragua 1,786,510,378 1.46% 10 Indonesia 1,610,864,742 1.31% 11 Germany 1,184,468,535 0.97% 12 United Kingdom 830,654,860 0.68% 13 Vietnam 747,231,546 0.61% 14 France 745,184,875 0.61% 15 South Korea 667,922,357 0.54% Summary Statistics Variable Label N Mean Std Dev Minimum Maximum P1 Mexico 95 3.26 1.7 1.5 8.9 P2 AUTO 95 5.84 2.0 4.1 16.1 P3 ROW 95 6.16 1.1 4.7 13.4 Q1 Mexico 95 876,868,172 351,064,644 217,062,020 1,581,805,566 Q2 AUTO 95 69,436,461 27,022,580 13,179,944 135,918,001 Q3 ROW 95 345,063,459 221,457,437 57,509,344 904,435,328 Pi Price, Qi Quantity 18 Table 1.2: 2011 Custom Value and Dutiable Percentage HTS 8512202040 (0% Tariff) HTS 8512204040 (2.5% Tariff) HTS 8544300000 (5% Tariff) Total 2011 Rank Mean Rank Country Custom Value Dutiable Percent Custom Value Dutiable Percent Custom Value Dutiable Percent Custom Value Dutiable Percent 1 1 Mexico 312,638,884 0.0% 129,845,258 11.4% 4,734,933,466 2.8% 5,177,417,608 2.9% 2 3 China 140,798,868 0.0% 56,445,377 100.0% 472,008,465 85.9% 669,252,710 69.0% 3 5 Taiwan 290,814,561 0.0% 98,475,182 100.0% 8,425,636 99.1% 397,715,379 26.9% 4 6 Honduras 0 0.0% 0 0.0% 384,234,329 0.7% 384,234,329 0.7% 5 9 Nicaragua 0 0.0% 0 0.0% 349,986,728 4.9% 349,986,728 4.9% 6 2 Philippines 0 0.0% 1,999,291 100.0% 332,920,817 99.7% 334,920,108 99.7% 7 4 Japan 135,336,694 0.0% 24,555,735 99.6% 66,276,230 97.5% 226,168,659 39.4% 8 8 Canada 74,864,146 0.0% 57,072,666 3.5% 45,866,392 12.8% 177,803,204 4.4% 9 13 Vietnam 0 0.0% 0 0.0% 170,623,130 100.0% 170,623,130 100.0% 10 10 Indonesia 0 0.0% 0 0.0% 169,148,836 100.0% 169,148,836 100.0% 11 15 South Korea 87,309,885 0.0% 12,516,727 100.0% 19,159,115 100.0% 118,985,727 26.6% 12 7 Thailand 171,836 0.0% 249,941 70.9% 115,289,909 44.9% 115,711,686 44.9% 13 11 Germany 79,113,767 0.0% 5,918,696 100.0% 22,960,323 86.5% 107,992,786 23.9% 14 14 France 6,459,393 0.0% 1,668,955 100.0% 50,303,908 11.4% 58,432,256 12.7% 15 12 United Kingdom 2,589,096 0.0% 0 0.0% 31,822,760 25.6% 34,411,856 25.9% Total (%CV Total) 1,130,097,130 (13.3) 388,747,828 (4.6) 6,973,960,044 (82.1) 8,492,805,002 HTS Code Description 8512.20.2040 Lighting equipment: For the vehicles of subheading 8701.20 or heading 8702, 8703, 8704, 8705 or 8711 0% 8512.20.4040 Visual signaling equipment: For the vehicles of subheading 8701.20 or heading 8702, 8703, 8704, 8705 or 8711 2.5% 8544.30.0000 Ignition wiring sets and other wiring sets of a kind used in vehicles, aircraft or ships 5% 19 Table 1.3: Model Restrictions Restriction Imposed None Homogeneity Model Null Wald ?2 Wald ?2 Synthetic Homogeneity 3.58 (0.1667) - Symmetry 6.00 (0.0143) 3.12 (0.0774) Rotterdam (?=0, ?=0) Homogeneity 1.70 (0.4268) - Symmetry 4.62 (0.0316) 3.66 (0.0559) CBS (?=1, ?=0) Homogeneity 3.43 (0.1799) - Symmetry 5.96 (0.0147) 3.20 (0.0735) NBR (?=0, ?=1) Homogeneity 2.60 (0.2724) . Symmetry 5.09 (0.0241) 3.69 (0.0547) AIDS (?=1, ?=1) Homogeneity 4.47 (0.1068) - Symmetry 6.48 (0.0109) 3.31 (0.0687) Unrestricted Synthetic None Null Specification1 Wald ?2 Synthetic (?=1) 0.32 (0.5740) Synthetic (?=0) 17.44 (<0.0001) Synthetic (?=1) 25.11 (<0.0001) Synthetic (?=0) 0.17 (0.6779) Rotterdam (?=0, ?=0) 17.64 (<0.0001) CBS (?=1, ?=0) 0.49 (0.7824) NBR (?=0, ?=1) 42.76 (<0.0001) AIDS (?=1, ?=1) 25.45 (<0.0001) 1 Homogeneity and Symmetry imposed, ? and ? are unrestricted in each test 20 Table 1.4: Model Coefficient Results N = 94 Synthetic Rotterdam CBS NBR AIDS Coeff. Est. Std Err Est. Std Err Est. Std Err Est. Std Err Est. Std Err ?1 -0.003 (0.013) 0.010 (0.013) -0.001 (0.012) 0.005 (0.015) -0.006 (0.014) ?3 -0.003 (0.012) -0.017 (0.013) -0.004 (0.011) -0.013 (0.014) -0.001 (0.013) ?1 -0.061 (0.151) 0.548 (0.042) * 0.021 (0.039) 0.546 (0.046) * 0.019 (0.043) ?2a -0.040 (0.024) 0.045 (0.014) * -0.029 (0.013) * 0.042 (0.015) * -0.032 (0.014) * ?3 -0.055 (0.116) 0.407 (0.040) * 0.008 (0.036) 0.412 (0.043) * 0.013 (0.040) ?11 -0.018 (0.048) 0.005 (0.013) 0.002 (0.012) 0.220 (0.014) * 0.216 (0.013) * ?12 0.012 (0.009) 0.009 (0.004) 0.008 (0.004) * -0.030 (0.005) * -0.030 (0.005) * ?13 0.006 (0.040) -0.014 (0.012) -0.010 (0.011) -0.190 (0.013) * -0.186 (0.012) * ?21a 0.012 (0.009) 0.009 (0.004) * 0.008 (0.004) * -0.030 (0.005) * -0.030 (0.005) * ?22a -0.014 (0.015) -0.007 (0.007) -0.007 (0.007) 0.065 (0.007) * 0.064 (0.007) * ?23a 0.002 (0.009) -0.002 (0.007) -0.001 (0.006) -0.035 (0.007) * -0.035 (0.007) * ?31 0.006 (0.040) -0.014 (0.012) -0.010 (0.011) -0.190 (0.013) * -0.186 (0.012) * ?32 0.002 (0.010) -0.002 (0.007) -0.001 (0.006) -0.035 (0.007) * -0.035 (0.007) * ?33 -0.008 (0.047) 0.016 (0.013) 0.011 (0.012) 0.225 (0.014) * 0.221 (0.013) * s1 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) s2 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) * 0.000 (0.000) s3 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) c1 0.001 (0.001) 0.000 (0.001) 0.000 (0.001) 0.000 (0.001) 0.001 (0.001) c3 0.001 (0.001) 0.002 (0.001) 0.001 (0.001) 0.002 (0.001) 0.001 (0.001) ? 1.156 (0.277) * 0 - 1 - 0 - 1 - ? -0.090 (0.218) 0 - 0 - 1 - 1 - R 2 Adj-R2 DW R2 Adj-R2 DW R2 Adj-R2 DW R2 Adj-R2 DW R2 Adj-R2 DW wd1 0.707 0.687 2.22 0.654 0.635 2.23 0.703 0.686 2.24 0.590 0.567 2.25 0.649 0.629 2.278 wd3 0.632 0.606 2.27 0.539 0.513 2.22 0.625 0.603 2.28 0.459 0.429 2.26 0.551 0.526 2.336 a Coeffiencent calculated from adding up restriction, ??i + ? = 1. 1 1 Mexico, 2 Auto Countries (Japan, Germany, South Korea), 3 Rest of World DW Durbin Watson, Critical Value = 1.71 * Statistically significant at 5% 21 Table 1.5: Compensated Elasticities Synthetic Rotterdam (?=0, ?=0) CBS (?=1, ?=0) NBR (?=0, ?=1) AIDS (?=1, ?=1) 1992 Share Compensated Price Elasticity ?11 0.005 (0.021) 0.008 (0.021) 0.003 (0.020) -0.025 (0.023) -0.030 (0.022) ?12 0.010 (0.012) 0.014 (0.007) * 0.014 (0.007) * 0.057 (0.007) * 0.057 (0.007) * ?13 -0.015 (0.018) -0.022 (0.020) -0.016 (0.018) -0.032 (0.021) -0.026 (0.020) ?21 0.057 (0.068) 0.083 (0.041) * 0.080 (0.040) * 0.337 (0.044) * 0.335 (0.043) * ?22 -0.050 (0.078) -0.066 (0.067) -0.069 (0.064) -0.274 (0.072) * -0.280 (0.069) * ?23 -0.006 (0.060) -0.016 (0.063) -0.011 (0.059) -0.063 (0.067) -0.055 (0.064) ?31 -0.033 (0.041) -0.051 (0.045) -0.037 (0.041) -0.072 (0.048) -0.059 (0.044) ?32 -0.002 (0.023) -0.006 (0.024) -0.004 (0.023) -0.024 (0.026) -0.021 (0.025) ?33 0.035 (0.044) 0.057 (0.048) 0.041 (0.043) 0.096 (0.051) 0.080 (0.047) Mean Share Compensated Price Elasticity ?11 0.008 (0.028) 0.010 (0.025) 0.003 (0.024) -0.057 (0.027) * -0.063 (0.026) * ?12 0.015 (0.008) 0.017 (0.008) 0.016 (0.008) * 0.022 (0.009) * 0.022 (0.009) * ?13 -0.024 (0.025) -0.026 (0.023) -0.019 (0.021) 0.034 (0.025) 0.041 (0.023) ?21 0.102 (0.054) 0.109 (0.055) 0.106 (0.053) * 0.147 (0.059) * 0.146 (0.057) * ?22 -0.091 (0.084) -0.088 (0.089) -0.091 (0.084) -0.099 (0.095) -0.107 (0.091) ?23 -0.011 (0.078) -0.022 (0.083) -0.015 (0.079) -0.048 (0.088) -0.039 (0.085) ?31 -0.031 (0.033) -0.035 (0.031) -0.026 (0.028) 0.045 (0.033) 0.054 (0.031) ?32 -0.002 (0.016) -0.004 (0.017) -0.003 (0.016) -0.010 (0.018) -0.008 (0.017) ?33 0.034 (0.034) 0.039 (0.033) 0.029 (0.030) -0.036 (0.036) -0.047 (0.032) 2012 Share Compensated Price Elasticity ?11 0.008 (0.035) 0.014 (0.035) 0.005 (0.032) -0.041 (0.038) -0.050 (0.035) ?12 0.025 (0.013) 0.023 (0.011) * 0.022 (0.011) * -0.009 (0.012) -0.009 (0.012) ?13 -0.033 (0.035) -0.037 (0.032) -0.027 (0.029) 0.050 (0.035) 0.059 (0.032) ?21 0.135 (0.071) 0.124 (0.062) * 0.121 (0.060) * -0.048 (0.067) -0.051 (0.065) ?22 -0.113 (0.098) -0.100 (0.101) -0.104 (0.096) 0.003 (0.108) -0.005 (0.103) ?23 -0.022 (0.090) -0.025 (0.094) -0.017 (0.089) 0.045 (0.100) 0.056 (0.096) ?31 -0.023 (0.024) -0.025 (0.022) -0.018 (0.020) 0.035 (0.024) 0.041 (0.022) ?32 -0.003 (0.011) -0.003 (0.012) -0.002 (0.011) 0.006 (0.013) 0.007 (0.012) ?33 0.026 (0.026) 0.028 (0.024) 0.021 (0.021) -0.040 (0.026) -0.048 (0.023) * 22 Table 1.6: Uncompensated Elasticities Synthetic Rotterdam (?=0, ?=0) CBS (?=1, ?=0) NBR (?=0, ?=1) AIDS (?=1, ?=1) 1992 Share Marshallian Price Elasticity ?11 -0.652 (0.052) * -0.539 (0.047) * -0.639 (0.044) * -0.571 (0.052) * -0.671 (0.048) * ?12 -0.101 (0.014) * -0.078 (0.010) * -0.095 (0.009) * -0.035 (0.011) * -0.051 (0.010) * ?13 -0.304 (0.027) * -0.264 (0.027) * -0.300 (0.025) -0.272 (0.030) * -0.308 (0.027) * ?21 -0.425 (0.119) * -0.185 (0.093) * -0.371 (0.091) * 0.085 (0.100) -0.099 (0.097) * ?22 -0.132 (0.079) -0.111 (0.068) -0.145 (0.064) * -0.317 (0.072) * -0.353 (0.069) * ?23 -0.219 (0.072) * -0.135 (0.073) -0.211 (0.069) * -0.174 (0.078) * -0.247 (0.075) * ?31 -0.626 (0.122) * -0.974 (0.101) * -0.676 (0.091) * -1.006 (0.109) * -0.709 (0.100) * ?32 -0.102 (0.030) * -0.162 (0.028) * -0.112 (0.026) * -0.182 (0.030) * -0.131 (0.029) * ?33 -0.317 (0.234) * -0.350 (0.063) * -0.240 (0.057) * -0.316 (0.068) * -0.206 (0.062) * Mean Share Marshallian Price Elasticity ?11 -0.536 (0.048) * -0.538 (0.049) * -0.541 (0.046) * -0.602 (0.054) * -0.606 (0.050) * ?12 -0.067 (0.010) * -0.066 (0.010) * -0.066 (0.010) * -0.060 (0.011) * -0.060 (0.011) * ?13 -0.436 (0.039) * -0.442 (0.040) * -0.432 (0.037) * -0.379 (0.044) * -0.370 (0.040) * ?21 -0.240 (0.106) * -0.190 (0.109) -0.229 (0.105) * -0.133 (0.117) -0.170 (0.113) ?22 -0.142 (0.084) -0.133 (0.089) -0.142 (0.084) -0.141 (0.095) -0.155 (0.091) ?23 -0.270 (0.103) * -0.248 (0.108) * -0.269 (0.104) * -0.261 (0.116) * -0.278 (0.112) * ?31 -0.564 (0.057) * -0.572 (0.061) * -0.560 (0.055) * -0.498 (0.066) * -0.486 (0.060) * ?32 -0.083 (0.017) * -0.086 (0.018) * -0.084 (0.017) * -0.092 (0.020) * -0.089 (0.019) * ?33 -0.370 (0.050) * -0.368 (0.052) * -0.376 (0.047) * -0.448 (0.056) * -0.456 (0.052) * 2012 Share Marshallian Price Elasticity ?11 -0.369 (0.065) * -0.534 (0.055) * -0.395 (0.051) * -0.587 (0.060) * -0.448 (0.055) * ?12 -0.044 (0.017) * -0.078 (0.014) * -0.051 (0.013) * -0.109 (0.015) * -0.082 (0.014) * ?13 -0.582 (0.091) * -0.832 (0.070) * -0.608 (0.065) * -0.743 (0.076) * -0.519 (0.070) * ?21 -0.086 (0.103) -0.122 (0.100) -0.103 (0.097) -0.279 (0.107) * -0.258 (0.104) * ?22 -0.154 (0.098) -0.145 (0.101) -0.145 (0.096) -0.039 (0.108) -0.044 (0.103) ?23 -0.344 (0.141) * -0.382 (0.144) * -0.342 (0.139) * -0.290 (0.154) -0.246 (0.149) ?31 -0.423 (0.044) * -0.305 (0.035) * -0.403 (0.032) * -0.249 (0.038) * -0.347 (0.035) * ?32 -0.076 (0.013) * -0.055 (0.013) * -0.073 (0.012) * -0.046 (0.014) * -0.064 (0.013) * ?33 -0.556 (0.063) * -0.379 (0.047) * -0.538 (0.042) * -0.452 (0.051) * -0.612 (0.046) * 23 Table 1.7: Expenditure Elasticities Synthetic Rotterdam (?=0,?=0) CBS (?=1, ?=0) NBR (?=0, ?=1) AIDS (?=1, ?=1) 1992 Share Expenditure Elasticity ?1 1.058 (0.076) * 0.882 (0.068) * 1.034 (0.063) * 0.878 (0.074) * 1.030 (0.069) * ?2 0.775 (0.149) * 0.431 (0.130) * 0.727 (0.126) * 0.405 (0.139) * 0.699 (0.135) * ?3 0.955 (0.182) * 1.486 (0.146) * 1.028 (0.132) * 1.504 (0.158) * 1.046 (0.144) * Mean Share Expenditure Elasticity ?1 1.039 (0.075) * 1.045 (0.081) 1.040 (0.075) * 1.041 (0.088) * 1.036 (0.082) * ?2 0.653 (0.167) * 0.571 (0.172) 0.639 (0.166) * 0.536 (0.184) * 0.602 (0.179) * ?3 1.017 (0.090) * 1.026 (0.101) 1.019 (0.091) * 1.038 (0.109) * 1.032 (0.100) * 2012 Share Expenditure Elasticity ?1 0.995 (0.149) * 1.444 (0.112) * 1.055 (0.104) * 1.439 (0.122) * 1.050 (0.113) * ?2 0.584 (0.190) * 0.648 (0.195) * 0.590 (0.189) * 0.609 (0.209) * 0.548 (0.203) * ?3 1.056 (0.100) * 0.739 (0.073) * 1.014 (0.066) * 0.748 (0.079) * 1.023 (0.072) * 24 Table 1.8: Tariff Removal Incidence Estimates Elasticity Price Wedge Nominal Change ?K ?A PK PA PK PA 0.071 -0.142 67% 33% - $0.74 - $0.36 0.142 -0.142 50% 50% - $0.55 - $0.55 0.213 -0.142 40% 60% - $0.44 - $0.66 1.0 -0.142 12% 88% - $0.13 - $0.97 Table 1.9: Year Ending Q3 Wire Harness Exports to United States 2011 HTS Code South Korea Japan Germany 0% Tariff 8512202040 63,529,298 72.8% 96,167,398 59.7% 66,301,813 74.7% 2.5% Tariff 8512204040 9,164,612 10.5% 17,942,114 11.1% 4,402,778 5.0% 5.0%Tariff 8544300000 14,562,706 16.7% 46,975,155 29.2% 18,038,652 20.3% Total 87,256,616 161,084,667 88,743,243 2012 HTS Code South Korea Japan Germany 0% 8512202040 80,788,951 71.2% 135,612,948 61.5% 35,845,601 68.2% 2.5% 8512204040 13,176,928 11.6% 28,120,821 12.8% 4,126,761 7.9% 5.0% 8544300000 19,529,538 17.2% 56,663,008 25.7% 12,557,182 23.9% Total 113,495,417 220,396,777 52,529,544 25 Figure 1.1: Wire Harness Imports 1989-2012 (Custom Value) Figure 1.2: Wire Harness Total Expenditure (In $Millions) 0 10 20 30 40 50 60 70 80 90 Bil lion s 0 10 20 30 40 50 60 70 80 90 100 1989 1989 1990 1991 1992 1992 1993 1994 1995 1995 1996 1997 1998 1998 1999 2000 2001 2001 2002 2003 2004 2004 2005 2006 2007 2007 2008 2009 2010 2010 2011 2012 26 Figure 1.3: Wire Harness Market Share (Percent of Total Expenditure) Figure 1.4: Wire Harness Quantity (In Billion Units) 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 1989 1991 1994 1996 1999 2001 2004 2006 2009 2011 Mexico Auto ROW 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 1989 1991 1994 1996 1999 2001 2004 2006 2009 2011 MexQ AutoQ RowQ 27 Figure 1.5: Wire Harness Price (USD per 100 Custom Value Units) Figure 1.7: Change in Welfare $0 $3 $6 $9 $12 $15 $18 1989 1991 1994 1996 1999 2001 2004 2006 2009 2011 MexP AutoP RowP P Q S 0 P 0 D P os t - KO R US w el far e change : U.S. Importe r: $432,000 S .K Ex porte r: $456,000 T otal: $888,000 1 19 Mi ll ion Q 0 P 0 = $4.10 100 uni ts S 1 P K = $3.74 100 uni ts P K Q 1 121 Mi ll ion P = $3.36 100 uni ts P = 0.071, = - 0.142 U.S. Importe r Surplus Cha ng e S .K. E x porte r Surplus C ha ng e 28 Chapter 2 Logistics Performance and Tariffs in the U.S. Apparel Import Market 2.1 Introduction The objective of this study is to measure the effects of tariffs as compared to export country logistics and transportation in the United States apparel import market in order to gain insight on Haitian economic development. An OLS model, zero-inflated Poisson (ZIP) and Heckman Two- Step Selection model are used to analyze the impacts of tariffs and logistics on the potential of apparel products from Haiti to the United States. Logistics will be the primary focus of this study, with the assumption tariff relief will not have a positive impact if the product cannot be moved to the export market. Analysis is conducted in two phases. First, the U.S. apparel import market in general. Results from the general import market indicate tariffs do not affect import demand. In order to uncover the market activity hidden from general market analysis, the second phase of the study analyses imports for apparel made of cotton, since the U.S. is a major exporter of cotton to the world. First, an ordinary least squares (OLS) model is used for initial evaluations of variable influences on imports. The OLS model presents a trade-off of two undesirable conditions of model assumption violation and an analytical problem, which leads to the use of the zero-inflated Poisson and the two-step Heckman selection models. 2.1.1 Haiti The United States has offered assistance beyond program funding to Haiti over the past few decades. On May 5th, 2010, the United States Congress approved House Resolution 5160, also known as the Haitian Economic Lift Program, or HELP. H.R. 5160 allowed the free trade of 29 several apparel products from Haiti to the U.S. market through the extension of the Caribbean Basin Economic Recovery Act of 1983, or CBERA, through September 2020. Haiti has an existing apparel industry from which it can build. Three actions from the United States have occurred in an effort to help Haiti grow to a more developed country. Two of the actions are directed through the apparel industry. The first was the original CBERA legislation from 1983 removing the U.S. tariff on Haitian imports, which has been continued through the HELP legislation. The second was the more recent action of building the Haitian Apparel Center. The third, more general action is the focus on infrastructure improvements, such as, road developments and water management. This study focuses on the first and third actions with the intent to provide information for use by procurement managers and other decision makers who potentially would actually allocate resources toward the apparel industry. 2.2 Literature Review Several studies are available regarding trade and its impact from tariffs and logistics. Baier and Bergstand (2006) used a gravity model to determine if free trade agreements actually increase the members? trade. Their study is focused on the endogeneity bias resulting from selection of countries to participate in the free trade agreement. They pointed out ?free trade agreement? as a RHS variable is often found to be an inconsistent, and sometimes, an endogenous variable which is negatively correlated to the error term. Baier and Bergstrand determined free trade agreements double trade between the participant countries after 10 years. The problem of selection bias is common in trade literature. Vandenbussche et al. (1999) used the Heckman Two-Step (Heckman) model to analyze trade flows of imports from countries named in European anti-dumping suits versus those which were not named as countries which 30 dump products. They found anti-dumping policies can have adverse or positive outcomes on welfare for the enforcing country depending on the commodity to which they are applied. Other two-step models found in the literature involve count models. Burger et al. (2009) use the logit model to select observations for Poisson and negative binomial count models to address problems of selection bias often found in gravity models. Literature with respect to transportation costs and infrastructure are abundant. An enduring finding in the literature is landlocked countries are at a disadvantage in transport costs to non- landlocked countries. This disadvantage is often increased for landlocked countries with low infrastructure capabilities. Lim?o and Venables (2001) developed a transport cost factor from the ratio of cost, insurance and freight (CIF) to freight on board (FOB) of trade between countries. They found trade between countries had a cost elasticity of approximately -3.0 with this ratio. A similar ratio is used for this study by replacing freight on board with customs value. Guasch and Kogan (2001) looked at inventory levels of raw material and finished goods in developing countries sourced from more developed nations. Their article involved analyzing the difficulties firms face when attempting to take advantage of free trade legislation, such as that provided by the HELP legislation. Using data compiled during the 1970s and 1980s, the authors estimate inventories in relation to GDP are two to three times larger for countries operating in developing countries as those operating in the U.S. They pointed out these increased inventories are held at interest rates of 15-20 per cent in developing countries. One of the theories driving their paper involves the term ?stockout,? which refers to firms running out of finished product needed to meet unexpected demands. Also, it can apply to firms? disabilities to deliver product on time to meet scheduled demand. Both of these are drivers for companies to increase inventories, and thus increase their costs. Gausch and Kogan?s modeling approach is quite 31 simple as they only use an OLS regression to show higher inventories follow lower capable infrastructure systems. Also, they determined a single standard deviation decrease in infrastructure increased inventory costs up to 47%. This study uses a nontraditional version of the gravity model for analysis of unilateral trade effects in a particular industry, apparel, at a relatively low aggregation level. The model used in this study has features of the traditional gravity model, but should be considered an alternative form of it. The typical gravity model uses geopolitical country characteristic variables such as common language as indicators of trade frictions. Also, economic variables, such as GDP and GDP per capita are common in gravity models. This study?s model uses similar variables. The deviation from the traditional gravity model comes from the lack of multi-lateral trade considerations. Commodity and raw material types are of interest in this study, which leads to a need of substantially disaggregated data. Data sources for this level of disaggregation were found to be generally limited to developed countries. This type of limitation on data often results in studies involving relatively few countries (Chi and Kilduff 2010). A search for data from the World Trade Organization (WTO) limited to data aggregated to all clothes. Information from the UN Comtrade data set was limited to information at the 6-digit HTS level, which would result in a weighted average of tariff information and reduce unit tariff information. More importantly, Haiti is not an observed country in the UN Comtrade data for apparel commodities6. An additional intent of the study is to gain insight on country characteristics influence on if trade even occurs, let alone the value of the trade, which requires observations of zeros from several countries. Combined, restrictions of data result in some conflicts between the intent of 6 A search for data in the UN Comtrade database under HTS codes 610210, 610220, 610230 and 610290 did not have information on Haiti, http://comtrade.un.org/db/mr/daCommodities.aspxy 32 this study and the traditional gravity model. As a result, this study employs a structural model similar to the gravity model, but is not in line with the traditional gravity model found in trade literature (Anderson, 2011). 2.3 OLS to Selection Model Progression The gravity model is a commonly used instrument for trade analysis over the last several decades. Silva and Tenreyro (2006) credit its initial development to Nobel economist Jan Tinbergen. As Silva and Tenreyro explain, the basic concept of the gravity model is countries of large masses (such as GDPs) will be attracted to each other for trade, yet discouraged by the physical distance between them. Equation 2.1) shows the typical form: 2.1) , where Tij is the total exports plus imports between the importing country i, and the exporting country j. Gi and Gj represent the GDPs of the two countries. The geographic distance between the two countries is represented by Dij. The coefficient ?1 is the potential to generate trade for the importer, ?2 is the potential to attract trade for the exporter, and ?3 is a factor of trade deterred by the distance between countries. ?3 is expected to be negative theoretically, ceterus peribus. A common procedure is to take the natural log version of both sides of Equation 2.1) to estimate the model with OLS. Estimating the natural log of the dependent variable is often necessary in order to maintain basic OLS assumptions. Figure 2.1 shows the distribution of the error term when the assumption of homoscedasticity is violated with the non-logged dependent variable, but maintained with the logged version. The normality assumption is also maintained with the logged version of the OLS model and lost with the non-logged version for the data in this article. 33 For this study, the model estimates only imports of apparel into the United States from other countries under Chapter 61 of the Harmonized Tariff Schedule (HTS61) from the USITC which are called out in the HELP legislation. Taking the natural log of both sides of Equation 2.1) and then adding controls for country characteristics results in a mixed model of log-log and log-level coefficients: 2.2) Since the United States is the only importer, Tij becomes Iij to estimate the amount of imported HTS61 products by country i from country j. For this article, Gi and Ci are the GDP per capita and a vector of characteristics, respectively, for the importing country. Since the United States is the only importing country, Ci and Gi are essentially constants in the equation. Gj is the GDP per capita for the exporting country j. Cj represents a vector of characteristics of the exporting country. Dij represents the distance in kilometers from the United States and the exporting country. This import version of the gravity model is similar to that used by Alam et al. (2009) and Chi and Kilduff, which use time series as opposed to cross-sectional data in this study. A bit of a paradox is created with the logged version of the gravity model though. The natural log of zero is undefined, which removes any observations of zero imports from country j from the dataset. This condition leaves the analyst with a choice between selection bias or OLS assumption violations, as discussed in the model results section. 2.3.1 Zero-Inflated Count Model Burger et al. (2009) used count models to address the Log-OLS paradox from the zero trade conditions. Wooldridge found count models, such as the negative binomial, can be used for estimated non-negative continuous data (2002). The zero-inflated negative binomial model 34 consists of two processes. The first process of the ZINB model is to determine if country j has a probability of no trade. This is determined with a binary logit outcome, where values of 0.5 and greater result in a value of 1, indicating no probability of exports. Values of less than 0.5 result in a 0 indicating country j is likely to export to the United States. Equation 2.3) shows the logit prediction model estimating the probability of trade: 2.3) Since the exp(X??) is between 0 and 1 for any X?? < 0, Equation 2.3) is less than 0.50 for any negative coefficient. In other words, a negative coefficient signals an increased probability to export to the United States. The second part of the ZINB is the count model, a negative binomial count model. The negative binomial regression estimator is based on the Poisson regression distribution and model. As explained by Cameron and Trivedi (2009), the Poisson model is represented by: 2.4) ,2,3?. where, 2.5) Equation 2.5) is the intensity or frequency at which the count would occur in a given time. For this study, the count is the import dollar amount (customs value), which fits with Wooldridge?s finding of a continuous non-negative number. ?0 represents the constant proportion of the mass between the importing and exporting countries. ??Xij represents a vector of comparisons between the two countries along with their coefficients, as in the OLS model. The important condition for accuracy of the Poisson distribution is the equality of the mean and variance of the dependent variable Iij, or equidispersion, where V[Iij] = E[Iij] = ?. Over- 35 dispersion occurs when the variance exceeds the mean, V[Iij] > E[Iij], which leads to the excessive zero condition. The excessive zero condition refers to a greater amount of actual zeros in the data set than what the Poisson model estimates. This further leads to over estimates of trade because the zeros are under-represented. The negative binomial model includes an additional term to modify the variance to account for the over-dispersion condition of the Poisson, V[Iij] = ?ij + ?g(?ij). If the ? coefficient is significantly different from zero, the negative binomial model is preferable to the Poisson. Estimation results for ? in the present study indicated it to be not significantly different from zero, leading to use of a zero-inflated Poisson. 2.3.2 Heckman Selection The Heckman Selection model uses a probit regression analysis for prediction if the dependent variable will occur. An OLS regression simultaneously occurs using the probability the probit model predicted. A positive coefficient in the probit selecting model indicates the country is likely to export HTS61 products to the United States. Cameron and Trivedi (2009) explain the model as a method to deal with left-truncated observations which should have been censored: 2.6) In equation 2.6), ?i is an error term and ?1 is the coefficient estimate of the X1i variable vector from the probit regression in step 1 of the Heckman. This portion of the equation predicts occurrence of the event, not magnitude. ?2 is the estimator for the second-step OLS regression of the X2i variable vector. ? is the estimated inverse Mills ratio7. The key coefficient in the Heckman is ?12 which determines the correlation between the errors of the probit and OLS 7 The Inverse Mills Ratio is the ratio of the normal probability density function to the cumulative density function, 36 model. If it is statistically different from zero, then selection bias is present in the basic OLS regression without using the first-step probit selection estimator. The resulting estimation model for country j is: 2.7) lcv =?0 + ?1 perc + ?2unit +?3 lpi + ?4X + ?12?( ?0 + ?1 perc + ?2unit +?3lpi + ?4Z)+ ? This model is estimated in Table 2.4 where Xj and Zj represent a set of fixed effect country characteristics in the OLS regression model and probit selection model, respectively. 2.4 Data for Estimation The customs value and tariff rates of apparel commodities from each country exporting to the United States from 2007 to 2011 provide the independent and dependent variables for the first analysis phase. The unit of analysis is the exporting country. Customs value data comes from the USITC. Tariff data is available from the USITC?s Harmonized Tariff Schedule at the 8-digit code level, of Chapter 61 of the HTS (HTS61). These are described as knitted or crocheted articles of apparel and clothing accessories and limited to coats, suits, sweaters and t-shirts. The tariff data is broken into a unit rate (unit) and a percentage rate (perc). A common practice is to convert the unit tariff to a percentage. However, this would remove the variable from observation because it would be divided by zero in several cases. For this reason, the percentage and unit tariffs are left separated. HTS Chapter 61 contains several 8-digit codes. Table 2.1 shows the top 20 countries involved in the import market for articles of clothing from the remaining data set. China is by far the leader in exports to the United States, followed by Vietnam and Honduras, as determined by customs value. Haiti supplies roughly 2% of the U.S. import market. 37 Logistics related variables come from the World Bank?s Logistics Performance Index (LPI) components. The World Bank completed LPI assessments in 2006, 2009 and 2011, which were reported in 2007, 2010 and 2012 respectively. LPI ratings were linearly interpolated in order to fill in the missing years of data. Distance and other country characteristics come from the ?Gravity Dataset? available from the Institute for Research on the International Economy (Centre d'Etudes Prospectives et d'Informations Internationales ? CEPII). This information is important for consideration of countries? natural trade advantages and restrictions. Countries with distance not available in the CEPPI data set were estimated with consideration of their nearest geographical neighbor?s distance. Disdier and Marette (2010) make use of common language and colonial ties as fixed effects for use with the gravity model. This study uses similar. Finally, the USITC is also the source for tariff-related information. The total matrix of observations includes 41 HTS codes from 164 countries over eac of five years for a total of 33,620. Descriptive statistics and coefficient correlations are available in Table 2.2 and Table 2.3. 2.5 General Apparel Imports The OLS, ZINB and Heckman models are used progressively to review the impacts of logistic and tariff variables on a country?s apparel imports to the United States. The dependent variable is the natural log of the customs value of HTS61 imported products, lcv. Two versions of the model are available for review and initial assessment of data with OLS. Each model shows the influence of different treatments of the data and variable considerations. Apparel made of animal byproducts (mainly wool) and cotton are compared against the base material, 38 synthetic or man-made materials. Dummy variables for the observation year are added using 2007 as the base year. Each model indicates no change in the intercept for a given year. The first two columns of Table 2.4, OLS1 (truncated) and OLS2 (censored), show the effect the logged version of the dependent variable has on the results. Observations of zero trade (?zeros?) from a country result in the natural log of zero, which is undefined. This removes the observation from the data set and leaves only positive observations. The difference between the truncated models and the censored models is the undefined dependent variables which are replaced with zeros in the censored. This is easily seen with the large reduction in observations at the bottom of Table 2.4 in columns 1 and 2. The large change in intercept shows the upward bias present from only considering observations in which trades occurred. Two analytical concerns come from the results. First, the coefficient estimates can be exaggerated because they are over-influenced by the observations remaining in the restricted estimating model. Several of the variables change in magnitude and significance when including the zeros. The second analytical concern from the unobserved zeros is the reason for the lack of trade. Without the zero observations, all the analyst knows is trade did not occur. The lack of imports may have been due to the tariff rate or another cost condition, such as high labor rates. The understanding of cost variables is diminished without the zero observations. The OLS model permits a cursory understanding of the impacts of the variables in the model. Tinbergen?s original model assumed trade was discouraged by distance. The distance variable was initially found to be counter to this hypothesis indicating importers were encouraged to import more as distance increased. This is likely due to misspecification of the variable because it does not consider economies of size from ever increasing cargo ships. However, adding distexcon more accurately reflects an updated consideration for the cost of distance. The 39 distexcon variable is likely picking up some total factor productivity considerations from the exporting country. The percent of total imports variable trade is likely impacting similarly. The most notable result is the statistically significant backward percentage tariff. The first and second analytic concerns create a paradox with the log-linear gravity model. Replacing the undefined value of ln(0) allows for consideration of zero trades, but it also brings back violation of the homoscedastic assumption. Replacement of the undefined ln(0) with a small number, such as 0.5 or 1, is a common practice. However, as Burger et al. (2009) suggest, this is arbitrary and potentially misleading. Further, in this instance, the estimated residuals remain heteroskedastic. Estimating the model without the zeros addresses the homoscedastic assumption concern, but removes the ability to fully assess the effects of the independent variables. This leads the use of the zero- inflated count model. The OLS gravity model estimates are of concern for analysis, but provide a baseline for the count model estimates. Column 3 of Table 2.4 shows the results of the ZIP, which results from a statistically insignificant ? estimate. Interpretation of the ZIP model is done with the inflated logit and the Poisson count model. In summary, the ZIP model confirms the direction of the OLS results. The percent of imports received in the U.S. from country j trade is still a strong determinate of whether the country will export HTS61. Increased ratios of CIF to CV, tradcost, decrease the likelihood of importing from the country and the amount of the imports if the decision is made to do so. The tariff variables show mixed results, which are hinted toward in the OLS models. The logit model estimates an increase in the percent tariff will decrease the odds of having zero trade. Clearly, this is a problematic result of the model. The unit tariff coefficient is in the theoretically correct direction for predicting probability of trade. 40 The lpi coefficient is also in the unanticipated direction and statistically different from zero for predicting the odds of exporting to the United States. However, other logistics related measures are in the expected direction. The exdocs variable indicates additional documents necessary to arrange export will reduce the odds of export occurring. The distexcon variable has an effect on the probability of imports, and is influential on the amount of imports. Landlocked countries and countries with a English as a common language apparently have no disadvantage nor advantage, respectively. The Heckman model in Column 4 of Table 2.4 uses the same independent variables as the ZIP. As for the tariff variables, they still show a similar condition as with the ZIP models. The percentage tariff coefficient perc is shown to be significant at the 1% probability level in the backward direction in probit part of the Heckman, but in the theoretically correct direction in the OLS portion. The unit tariff shows the same expected direction as well. The overall results from Table 2.4 shows percentage tariffs apparently increase imports of HTS61 products. The Heckman results in Table 2.4 are mostly consistent with the ZIP results. The Zero-Inflated Poisson and the Heckman models report similar coefficient estimates, indicating robustness to the results. Table 2.5 repeats estimation for the general apparel import market using the Heckman model. Table 2.5 further indicates robustness of the results and implications of estimating materials concurrently. The percent and the unit tariff variables in this study were estimated in the unexpected direction by each model. This would indicate tariff reduction would hurt a country?s chances of exporting to HTS61 products. This may be a case of endogeneity bias causing a ?backward? coefficient, as noted previously from the Baier and Bergstand (2006) paper. The tariff variables 41 were tested with the Durbin-Wu-Hausman8 test and found to show signs of endogeneity in the general apparel data. The endogenous condition of the tariff variables is likely due to the aggregation level of the apparel import data. The data set includes imports made from material of cotton, man-made or synthetic materials, and animal-based products, most of which are wool. The U.S. does not impose a unit tariff on cotton products. A review of HTS Chapter 61 shows the U.S. only directs unit tariffs (assessed per kilogram) toward wool-based products. Considering the U.S. is major exporter to the world for cotton, a unit tariff on cotton would theoretically hurt its own cotton industry. Segmenting the data according to material can help uncover the actual reactions to tariffs, and by how much a country?s LPI can override these reactions. This leads to analysis of the cotton apparel imports. 2.6 Cotton Apparel Imports Analysis of the cotton import market is done with import data from the same sources noted previously. Summary statistics are available in Table 2.6 for cotton apparel imports and compare similarly with those for the general import data. Table 2.8 displays results for estimations in the same format as Table 2.4, but for only the cotton apparel imports. As a result, unit tariffs are not in the data set. The most notable change from segmenting out cotton-based apparel is the change in direction of the tariff variable. Also, the endogeneous condition of the percent tariff goes away when only considering cotton-based apparel. Using the percent tariff coefficient estimate from the Heckman model, a 1% decrease in the tariff rate results in a 0.23%9 increase in expected average export custom value, should export occur. 8 Test results in Appendix B 9 (e-0.26) ? 1 = -0.229 42 Focusing on the infrastructure and transportation related variables, the logistics performance index is apparently not influential in the customs value an exporting country should expect. However, some variables related to the process of moving material from the exporting country to the U.S. are influential, but not all. The lpi and exdays variables are essentially a measure of performance of moving product within the exporting country. These variables are estimated as statistically no different from zero. This is an indication the U.S. importer is not concerned with performance within the exporting country and may further indicate costs of inefficiency within the exporting country are not passed on to the U.S. importer, and would put downward pressure on wage rates. The number of days to move material from the producing facility to the exporting dock, exdays, would increase inventory holding costs on the exporter side of the market, for example. The other transport related variables in Table 2.8 are the exdocs and distexcon. These two variables should reflect an interaction of the U.S. importer with the exporting country. Export documents, such as customs clearance and port authority related documents, would presumably require time from the U.S. importer. Additionally, the distexcon variable would be a measure of another inventory cost while the product is held on a cargo container. Depending on individual contract terms, the U.S. importer may take possession when the material leaves the exporting country?s dock, thus the distexcon directly influences the importer?s inventory turns. Considering the global trade of apparel, the U.S. importer has several options and may have a relatively strong elastic demand curve relative to supply, in particular to that of Haiti. As such, cost reductions from improvements within the exporting country?s infrastructure and transportation system would be kept within the exporting country. 43 Table 2.9 repeats the Heckman model for each observation year in the same manner as Table 2.5. The overall results for Table 2.8 and Table 2.9 remain the same as those from Table 2.5, with some exceptions. Most importantly for this study, the tariff variable condition maintains the expected direction and a similar magnitude. In the cotton-based apparel market, removal of a tariff can thus be better understood than compared to tariff relief in the general material market. As for transportation related variables, the ldist variable becomes statistically not different from zero, but the distexcon variable remains statistically significant for estimating the magnitude of export to the United States. The overall result from this condition may be further indication distance alone is not the same discouragement to trade it was when Tinbergen developed the gravity model. The cost of the distance needs to be considered directly as well. Table 2.10 displays repeated results for the same model specification displayed in Table 2.9, except for animal-based material apparel only, mostly wool. The overall results are much different from those of the cotton apparel. The percent tariff variable is not statistically significant as in the cotton market. Also, the distexcon variable is not significant. Comparing Table 2.9 to Table 2.10 shows cost considerations in the cotton apparel market do not necessarily flow into the wool apparel market. 2.7 Concluding Analysis Assessing the value of tariff relief through the 2010 HELP legislation in this study is limited to imported apparel made from cotton. Table 2.11 displays the top 10 sources of cotton-based apparel for the 2007-2011 period. China again is the market leader, but with less of the overall market compared to the general apparel market. Removal of the tariff gives Haiti an important advantage over established low-wage source countries with large economies, like China. 44 Removal of the tariffs for Haiti is advantageous for Haiti to lure market share from China and maybe Honduras, though Honduras has a zero tariff rate on most apparel products though the Central American Free Trade Agreement (CAFTA). However, countries like Mexico and Canada have zero tariffs, and better performing logistics indices. And moving forward, Haiti?s free trade status must be re-approved in 2020. This results in a potential for a decrease in probability of trade if the 1983 CBERA legislation is not re-approved, continuing to put much of Haiti?s fate in the hands of U.S. politicians. Haiti has some opportunities in the cotton apparel market through improvements to its transportation system. The interactive variable distexcon from Table 2.8, which measures the ratio of distance to export container cost, indicates a 1.4% increase in custom value for a 10% increase in the ratio, provided trade occurs. Increasing the distexcon ratio will also increase the odds of exporting to the U.S. Table 2.12 provides the average export container costs and marginal effects of the distexcon variable from the probit selection model within the Heckman model for Haiti, compared to China and Honduras. The benefit to Haiti for an incremental increase in the ratio is more than double than what it is for China. None of the countries can do anything about the distance, but Honduras and Haiti have more room to work on the container cost than China. If Haiti were to reduce container costs to $990, roughly 5%, it would increase its probability of exporting by 4%. The incremental benefits would reduce thereafter, but would continue to provide greater exposure to other cotton apparel commodities for Haiti. However, at the same time, China?s ratio would increase at a greater percentage from the same nominal decrease in container cost. Percentage decreases in container costs would provide more benefit to Haiti. 45 Table 2.13 displays a breakdown of the types of cotton apparel the U.S. imports from China, Honduras and Haiti for 2007-2011. Nearly all of Haiti?s cotton apparel exports to the U.S. were t-shirts. China exported over $4.5 billion dollars in cotton coats and sweaters, which likely have much more room for wage growth. Table 2.14 provides regression results for the percent tariff and the distexcon ratio from the same Heckman specified model ran for the general and cotton apparel markets10. The estimates are further segmented to coats, sweaters and t-shirts. It also has marginal effects for the percentage tariff and distexcon variables for the probability of exporting to the United States from the probit model within the Heckman. Haiti has significant advantage in tariff relief compared to China in the cotton coat market. Since these are percentage tariffs based on customs value, it is reasonable Haiti would have considerable advantage in any percentage cost reduction compared to China. This condition does not flow into the sweater market. And apparently, Haiti has peaked for increased probability of exporting t-shirts to the U.S. through tariff relief. The relationship to China is opposite for what it is in the coats market for tariff marginal effects. Section 2.8 Summary and Implications of Findings The U.S. Congress has continued efforts toward Haitian economic development through support of infrastructure improvements and tariff removal on imported apparel from Haiti. Assessment of the efforts is conducted with a modified unilateral version of the gravity model. Regression results indicate the aggregation level of the imported product is quite influential for model specification. The differences between the general material and cotton material results indicate some underlying market activities based on commodity type. Most assessments of trade conditions are done with highly aggregated data. This study does not go in to depth of the 10 Full model results available in Appendix B 46 underlying activities on the supply side of the market, but does suggest assumptions of the aggregated data may not be consistent at various aggregation levels on the demand side. In summary, Haiti apparently has some opportunities with cotton apparel from the removal of tariffs in the U.S. import market. The tariff removal provides opportunities, but the favorable conditions are time bound and out of control of Haiti. They are also limited to the production of cotton coats. Continued growth through exports of cotton apparel to the U.S. should be expected to come from reductions of export costs. A subtle finding from this study may be logistic performance and capabilities within the exporting country are not considered by U.S. importers. As a result, cost reductions from logistic improvements within Haiti can apparently be kept within the Haitian economy. A more apparent result from this study is U.S. apparel importers act differently depending on the material type. In other words, tariff removal on cotton apparel does not necessarily have the same benefit to wool apparel products. And within cotton apparel coats may have more opportunities for growth than sweaters and certainly more than t-shirts. The consideration is important for companies expecting greater access to the U.S. apparel market. Cotton apparel producers expecting growth via the HELP legislation may need to consider producing coats, not t-shirts. The consideration is also important for Haitian and U.S. officials using apparel as an instrument for Haitian economic development. Former U.S. President Clinton has acknowledged on multiple occasions a program he set up while Governor of Arkansas to provide low cost, or free, rice to Haiti while providing Arkansas farmers a channel to reduce excess supply (Gates and Clinton 2010) (O'Conner 2013). The program was intended to help U.S. farmers while 47 helping malnourished Haitian. The unintended consequence was crushing the Haitian rice farmer market. The U.S. has had the Earned Import Allowance Program (EIAP) with Haiti since 2006 through the Haitian Hemispheric Opportunity through Partnership Encouragement Act (HOPE), passed by the same committee that passed the HELP legislation. The program provides Haitian producers to have tariff-free export of 1 square meter equivalent (SME) of apparel to the United States for every 2 SMEs of material imported from the United States (Evans 2012). This essentially encourages Haiti to import U.S. cotton at a discount, which benefits U.S. cotton producers and provides less expensive exports of non-U.S. raw material for Haitian apparel producers, for example material from China. The program had very little impact the first 5 years with only $350,000 worth of credits claimed from January through August 2011. However, credits jumped to $18 million for the same time period in 201211. Haiti is estimated to supply 3.6% of the U.S. cotton apparel from this study?s data set, which represents nearly $200 million in custom value per year from 2007 to 2011, or roughly 2.5% of Haiti?s annual GDP. Cotton is also used in several other apparel types not considered in this study, so the comparison to Haiti?s GDP is conservative. The unintended consequence here may be the market destruction of the market for Haitian domestic cotton supply. Haiti used to have a cotton industry supplying export material to Europe. The well-known erosion and other environmental and economic conditions caused near elimination of Haiti?s cotton production by 1990. Expected benefits for HELP legislation should probably be tapered by the effects of the EIAP program. Employment growth may be horizontal from apparel material, but not vertical. 11 The timing coincides with entrance of South Korean apparel producer SAE A Trading Company to the Carocol Industrial Park on Haiti?s northern shore. 48 Table 2.1: USA Import Market for HTS Chapter 61 Apparel (in Millions) Country Total 2007 2008 2009 2010 2011 China $32,144 31.9% $5,624.51 27.4% $5,595.13 27.6% $6,045.76 33.9% $7,230.56 35.7% $7,647.74 34.7% Vietnam $7,903 7.8% $1,154.34 5.6% $1,504.94 7.4% $1,561.36 8.8% $1,766.39 8.7% $1,915.75 8.7% Honduras $6,900 6.8% $1,405.69 6.8% $1,509.67 7.4% $1,082.58 6.1% $1,327.91 6.6% $1,573.96 7.1% Indonesia $5,832 5.8% $868.62 4.2% $1,027.75 5.1% $1,144.19 6.4% $1,319.40 6.5% $1,472.35 6.7% Mexico $4,572 4.5% $1,082.00 5.3% $1,005.36 5.0% $780.95 4.4% $826.64 4.1% $877.17 4.0% El Salvador $4,200 4.2% $818.46 4.0% $883.63 4.4% $716.26 4.0% $886.86 4.4% $895.09 4.1% Guatemala $3,597 3.6% $762.15 3.7% $720.95 3.6% $670.47 3.8% $683.95 3.4% $759.44 3.4% Cambodia $3,569 3.5% $756.04 3.7% $752.25 3.7% $575.97 3.2% $683.26 3.4% $801.53 3.6% India $3,117 3.1% $599.51 2.9% $608.24 3.0% $556.10 3.1% $661.27 3.3% $691.60 3.1% Nicaragua $2,877 2.9% $452.00 2.2% $498.28 2.5% $534.01 3.0% $596.56 2.9% $796.08 3.6% Pakistan $2,655 2.6% $513.70 2.5% $526.45 2.6% $485.11 2.7% $532.80 2.6% $597.41 2.7% Bangladesh $2,278 2.3% $398.63 1.9% $475.97 2.3% $420.05 2.4% $506.74 2.5% $476.37 2.2% Haiti $2,019 2.0% $371.39 1.8% $328.56 1.6% $394.99 2.2% $396.95 2.0% $526.75 2.4% Peru $1,979 2.0% $452.07 2.2% $418.25 2.1% $346.14 1.9% $365.25 1.8% $397.18 1.8% Jordan $1,945 1.9% $452.18 2.2% $401.45 2.0% $342.92 1.9% $370.59 1.8% $377.96 1.7% Hong Kong $1,825 1.8% $896.46 4.4% $743.13 3.7% $91.52 0.5% $51.11 0.3% $42.49 0.2% Philippines $1,529 1.5% $464.26 2.3% $340.54 1.7% $239.75 1.3% $221.33 1.1% $262.71 1.2% Thailand $1,265 1.3% $338.58 1.6% $311.27 1.5% $193.00 1.1% $215.55 1.1% $206.19 0.9% Dom. Rep. $1,063 1.1% $189.53 0.9% $203.22 1.0% $169.11 0.9% $232.73 1.2% $268.48 1.2% Italy $949 0.9% $247.38 1.2% $217.18 1.1% $142.68 0.8% $153.06 0.8% $188.32 0.9% 49 Table 2.2: Descriptive Statistics for General Apparel Imports Variable Description Obs Mean Std. Dev. Min Max Source CV Custom value of apparel imports, in 1000 USD 33,558 3,007 44,326 0.0 3,402,162 USITC 8,445 11,950,363 87,756,925 0.3 3,402,162 lgdp Natural log of GDP 33,358 24.3 2.2 19.5 29.6 World Bank 8,445 25.9 2.0 19.7 29.6 lgdpop Natural log of GDP per Capita, proxy for wages 33,558 8.5 1.6 5.4 12.1 World Bank 8,445 9.0 1.4 5.5 12.1 tradcost Custom insurance and freight (CIF) for all commodities from exporting country divided by custom value of all product by export country, tradcost = [(CIF/CV)-1]*100 33,558 5.3 5.4 0.2 51 USITC 8,445 4.3 2.4 0.3 51 dist Distance in kilometers from exporting country to U.S. 22,187 8,791 3653 548 16,180 CEPII 8,143 8,590 3,747 548 16,180 excon Cost in USD for export of 20' sea container from exporting country 33,558 1,232 726 0 4,285 World Bank 8,445 1,015 456 0 4,285 distexcon Interactive variable: dist / excon 21,756 9.9 7.1 .3 36.9 8,029 10.5 7.6 0.3 36.9 perc Percent tariff on commodity 32,726 10.4 9.1 0 32 USITC 8,445 13.4 10.0 0.0 32 unit Unit tariff on commodity per kilogram 33,558 5.4 16.6 0.0 64.4 USITC 8,445 3.4 13.2 0 64.4 trade Percent of all U.S. imports from exporting country 33,348 0.6 2.2 0.0 19.2 USITC 8,428 1.6 3.7 0 19.2 exdays The median time, in days, for shipment from the product origin (factory) to the port of loading (dock, airport) 32,214 22.2 14.5 0 89 World Bank 8,166 16.4 9.1 5 89 Exdocs The number of documents required to export product 32,214 6.5 2.1 2.0 14.0 World Bank 8,166 5.8 2.0 2.0 13.0 lpi Logistical Performance Index Rating - The average of six performance measurables involving transportation and logistics for a country; based on survey results from the private sector 29,190 2.8 0.7 0.7 4.2 World Bank 8,166 3.1 0.6 0.7 4.2 Top line reflects entire data set. Bottom line reflects only those observations with CV > 0. 50 Table 2.3: Variable Correlation Matrix (Full Data Set) lcv lgdp lgdpop lpi tradcost dist excon distexcon perc unit trade exdays exdocs lcv 1.00 lgdp 0.39 1.00 <.0001 lgdpop 0.14 0.51 1.00 <.0001 <.0001 lpi 0.27 0.70 0.72 1.00 <.0001 <.0001 <.0001 tradcost -0.10 -0.34 -0.24 -0.19 1.00 <.0001 <.0001 <.0001 <.0001 dist -0.01 -0.10 -0.26 -0.10 0.14 1.00 0.27 <.0001 <.0001 <.0001 <.0001 excon -0.18 -0.16 -0.41 -0.36 -0.13 -0.06 1.00 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 distexcon 0.11 0.15 0.00 0.18 0.15 0.73 -0.57 1.00 <.0001 <.0001 0.96 <.0001 <.0001 <.0001 <.0001 perc 0.21 -0.04 -0.03 -0.03 0.03 0.14 0.05 0.04 1.00 <.0001 <.0001 <.0001 <.0001 0.00 <.0001 <.0001 <.0001 unit -0.08 -0.01 -0.01 -0.01 0.01 0.04 0.01 0.01 0.17 1.00 <.0001 0.01 0.02 0.03 0.22 0.22 0.14 0.25 <.0001 trade 0.29 0.45 0.18 0.30 -0.12 -0.11 -0.07 0.08 -0.09 -0.03 1.00 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 exdays -0.21 -0.37 -0.60 -0.54 0.05 0.27 0.75 -0.17 0.05 0.01 -0.16 1.00 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 0.01 <.0001 exdocs -0.19 -0.36 -0.63 -0.59 0.11 0.27 0.50 -0.08 0.05 0.01 -0.17 0.67 1.00 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 0.0239 <.0001 <.0001 51 Table 2.4: General Apparel Regression Results 1 OLS No Zeros 2 OLS with Zeros 3 Zero Inflated Count 4 Heckman Parameter OLS SE OLS SE Pois SE Logit SE OLS SE Probit SE Intercept 13.41 1.13 ** -5.45 1.20 ** 2.70 0.11 ** 4.43 0.56 ** 14.46 1.28 ** -2.63 0.33 ** lgdp 0.17 0.03 ** 0.84 0.03 ** 0.02 0.00 ** -0.39 0.02 ** 0.14 0.05 ** 0.23 0.01 ** lgdpop -0.55 0.06 ** -0.77 0.06 ** -0.05 0.01 ** 0.26 0.02 ** -0.45 0.06 ** -0.16 0.01 ** LPI -0.32 0.12 ** -0.17 0.12 -0.03 0.01 * 0.30 0.06 ** -0.31 0.12 * -0.02 0.03 ldist -0.25 0.12 * -0.63 0.13 ** -0.05 0.01 ** 0.42 0.06 ** -0.41 0.13 ** -0.25 0.04 ** distexcon 0.09 0.01 ** 0.06 0.01 ** 0.01 0.00 ** -0.01 0.00 ** 0.10 0.01 ** 0.01 0.00 ** Perc -0.08 0.01 ** 0.01 0.01 -0.01 0.00 ** -0.06 0.00 ** -0.09 0.01 ** 0.03 0.00 ** Perc2 0.01 0.00 ** 0.01 0.00 ** 0.00 0.00 0.00 0.00 0.01 0.00 ** 0.00 0.00 Unit 0.01 0.01 0.00 0.01 0.00 0.00 0.01 0.00 * 0.01 0.01 0.00 0.00 * Unit2 0.00 0.00 * 0.00 0.00 ** 0.00 0.00 0.00 0.00 0.00 0.00 * 0.00 0.00 tradcost -0.20 0.02 ** -0.05 0.02 ** -0.02 0.00 ** 0.02 0.01 -0.19 0.02 ** -0.01 0.01 lang -0.03 0.10 -0.28 0.10 ** 0.08 0.05 -0.05 (0.03) landlocked -0.69 0.13 ** 0.03 0.13 -0.08 0.06 0.04 (0.03) ExDays 0.05 0.01 ** 0.01 0.01 0.00 0.00 * 0.03 (0.01) ** ExDocs -0.16 0.03 ** -0.18 0.03 ** 0.06 0.01 ** -0.04 (0.01) ** Trade 0.15 0.01 ** 0.28 0.02 ** 0.01 0.00 ** -0.06 0.01 ** 0.13 (0.01) ** 0.04 (0.01) ** Animal 0.31 0.10 ** 0.49 0.09 ** 0.03 0.01 * 0.29 (0.10) ** Cotton 1.43 0.09 ** 3.05 0.11 ** 0.13 0.01 ** 1.42 (0.09) ** Year 2008 0.17 0.11 -0.01 0.12 0.02 0.01 0.19 (0.11) Year 2009 -0.20 0.11 -0.26 0.12 * -0.02 0.01 -0.17 (0.11) Year 2010 -0.10 0.11 -0.11 0.12 -0.01 0.01 -0.08 (0.11) Year 2011 -0.07 0.11 ** 0.02 0.12 -0.01 0.01 -0.07 (0.11) Obs 7,436 17,657 1,692 17,657 2,332 17,977 R2 0.2586 0.2941 Adj R2 0.2565 0.2933 AIC 57,302 57,627 Sigma 3.03 (0.03) ** Vuong/Rho 158.76** -0.15 ** *Statistically significant at 5%, ** Statistically significant at 1% for all Tables. 52 Table 2.5: General Apparel Repeated Heckman Results Parameter 2007 SE 2008 SE 2009 SE 2010 SE 2011 SE Intercept 14.47 2.90 ** 13.99 2.87 ** 3.38 3.24 14.97 2.82 ** 15.92 2.68 ** lgdp 0.20 0.10 * 0.03 0.10 0.76 0.10 ** 0.20 0.10 * 0.11 0.10 lgdpop -0.49 0.15 ** -0.44 0.12 ** -0.65 0.13 ** -0.29 0.15 -0.50 0.12 ** LPI -0.43 0.35 0.13 0.27 -0.63 0.30 * -0.97 0.39 * -0.28 0.23 ldist -0.45 0.28 -0.12 0.29 -1.24 0.32 ** -0.59 0.28 * -0.43 0.29 distexcon 0.08 0.02 ** 0.09 0.02 ** 0.13 0.02 ** 0.15 0.02 ** 0.10 0.02 ** Perc -0.07 0.02 ** -0.09 0.02 ** 0.00 0.02 -0.10 0.02 ** -0.09 0.02 ** perc2 0.01 0.00 ** 0.01 0.00 ** 0.01 0.00 ** 0.01 0.00 ** 0.01 0.00 ** Unit -0.01 0.03 -0.01 0.03 -0.03 0.03 0.05 0.02 * 0.05 0.03 unit2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 ** 0.00 0.00 * tradcost -0.20 0.05 ** -0.18 0.05 ** -0.30 0.06 ** -0.24 0.06 ** -0.11 0.04 * ExDays 0.02 0.02 0.04 0.02 * 0.05 0.02 ** 0.03 0.02 0.01 0.02 Trade 0.11 0.03 ** 0.11 0.03 ** 0.17 0.04 ** 0.12 0.03 ** 0.15 0.03 ** Animal 0.23 0.22 0.42 0.22 0.26 0.20 0.23 0.22 0.40 0.20 * Cotton 1.56 0.21 ** 1.40 0.21 ** 1.24 0.21 ** 1.28 0.21 ** 1.43 0.21 ** Select Intercept -2.26 0.76 ** -2.30 0.77 ** -2.68 0.73 ** -2.98 0.74 ** -2.33 0.74 ** Select lgdp 0.20 0.02 ** 0.22 0.02 ** 0.24 0.02 ** 0.23 0.02 ** 0.26 0.02 ** Select lgdpop -0.17 0.04 ** -0.15 0.03 ** -0.15 0.03 ** -0.21 0.04 ** -0.16 0.03 ** Select LPI 0.15 0.10 -0.05 0.07 -0.09 0.07 0.09 0.12 -0.10 0.06 Select ldist -0.26 0.08 ** -0.24 0.09 ** -0.27 0.08 ** -0.18 0.08 * -0.33 0.09 ** Select distexcon 0.01 0.01 * 0.01 0.01 0.02 0.01 ** 0.00 0.01 0.01 0.01 Select Perc 0.04 0.01 ** 0.03 0.01 ** 0.03 0.01 ** 0.03 0.01 ** 0.03 0.01 ** Select perc2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Select Unit 0.00 0.01 0.00 0.00 0.00 0.00 -0.01 0.00 ** 0.00 0.01 Select unit2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Select tradcost -0.02 0.02 -0.02 0.01 -0.03 0.01 * 0.01 0.01 0.02 0.01 Select lang -0.09 0.07 -0.07 0.06 -0.14 0.05 ** 0.04 0.07 0.01 0.06 Select landlocked 0.08 0.09 0.06 0.08 0.09 0.06 -0.01 0.08 -0.01 0.08 Select ExDocs -0.02 0.02 -0.05 0.02 * 0.00 0.02 -0.05 0.02 * -0.05 0.02 ** Select Trade 0.04 0.01 ** 0.04 0.01 ** 0.05 0.01 ** 0.04 0.01 ** 0.02 0.01 * N (OLS / Select) 335 / 3,571 501 / 3,654 536 / 3,654 353 / 3,528 609 / 3,570 AIC 11,759 11,650 11,312 11,276 11,747 53 Table 2.6: Descriptive Statistics for Cotton Apparel Imports Variable Description Obs Mean Std. Dev. Min Max Source CV Custom value of apparel imports, in 1000 USD 5,593 4,833 35,351 0.0 708,121 USITC 2,029 13,322 57,729 251 708,121 lgdp Natural log of GDP 5,593 24.3 2.2 19.5 29.6 World Bank 2,029 25.7 2.0 19.7 29.6 lgdpop Natural log of GDP per Capita, proxy for wages 5,593 8.5 1.6 5.4 12.1 World Bank 2,029 8.9 1.5 5.5 12.1 tradcost Custom insurance and freight (CIF) for all commodities from exporting country divided by custom value of all commodities from exporting country, tradcost = [(CIF/CV)-1]*100 5,593 5.3 5.4 0.2 51 USITC 2,029 4.3 2.2 0.3 26.1 dist Distance in kilometers from exporting country to U.S. 3,700 8792 3,652 548 16,180 CEPII 1,925 8,691 3,829 548 16,180 excon Cost in USD for export of 20' sea container from exporting country 5,593 1,232 726 0 4,285 World Bank 2,029 1,028 483 0 4,285 distexcon Interactive variable: dist / excon 3,626 9.9 7.1 .3 36.9 1,892 10.5 7.7 0.3 36.9 perc Percent tariff on commodity 5,593 10.1 5.4 0 16.5 USITC 2,029 10.5 6.3 0 16.5 trade Percent of all U.S. imports from exporting country 5,558 0.6 2.2 0.0 19.2 USITC 2,025 1.3 3.2 0 19.2 exdays The median time, in days, for shipment from the product origin (factory) to the port of loading (dock, airport) 5,369 22.2 14.5 0 89 World Bank 1,962 17.1 9.5 5 89 Exdocs The number of documents required to export product 5,369 6.5 2.1 2.0 14.0 World Bank 1,962 5.9 2.0 2 13 lpi Logistical Performance Index Rating - The average of six performance measurables involving transportation and logistics for a country; based on survey results from the private sector 4,865 2.8 0.7 0.7 4.2 World Bank 1,938 3.1 0.7 0.7 4.2 Top line reflects entire data set. Bottom line reflects only those observations with CV > 0. 54 Table 2.7: Cotton Variable Correlation Matrix (Full Data Set) lcv lgdp lgdpop lpi tradcost dist excon distexcon perc trade exdays exdocs lcv 1.00 lgdp 0.44 1.00 <.0001 lgdpop 0.11 0.51 1.00 <.0001 <.0001 lpi 0.28 0.70 0.72 1.00 <.0001 <.0001 <.0001 tradcost -0.13 -0.34 -0.24 -0.19 1.00 <.0001 <.0001 <.0001 <.0001 dist 0.04 -0.11 -0.26 -0.10 0.14 1.00 0.03 <.0001 <.0001 <.0001 <.0001 excon -0.23 -0.16 -0.41 -0.36 -0.13 -0.06 1.00 <.0001 <.0001 <.0001 <.0001 <.0001 0.00 distexcon 0.18 0.15 0.00 0.18 0.15 0.73 -0.57 1.00 <.0001 <.0001 0.98 <.0001 <.0001 <.0001 <.0001 perc 0.09 -0.06 -0.04 -0.03 0.04 0.22 0.08 0.06 1.00 <.0001 <.0001 <.0001 <.0001 0.02 0.00 <.0001 0.00 trade 0.28 0.45 0.18 0.30 -0.12 -0.11 -0.07 0.08 -0.15 1.00 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 exdays -0.24 -0.37 -0.60 -0.54 0.05 0.27 0.75 -0.17 0.08 -0.16 1.00 <.0001 <.0001 <.0001 <.0001 0.00 <.0001 <.0001 <.0001 <.0001 <.0001 exdocs -0.19 -0.36 -0.63 -0.59 0.11 0.27 0.50 -0.08 0.07 -0.17 0.67 1.00 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 55 Table 2.8: Cotton Apparel Regression Results OLS No Zeros OLS Zeros Zero Inflated Poisson Heckman Variable OLS 1 SE OLS 2 SE Count SE Logit SE OLS SE Probit SE Intercept 15.33 1.13 ** -2.36 1.22 2.19 0.25 ** 10.82 1.47 ** 12.50 2.64 ** -6.35 0.87 ** lgdp 0.15 0.03 ** 0.82 0.04 ** 0.04 0.01 ** -0.59 0.04 ** 0.16 0.08 0.34 0.02 ** lgdpop -0.58 0.06 ** -0.79 0.06 ** -0.07 0.01 ** 0.53 0.06 ** -0.54 0.12 ** -0.32 0.04 ** LPI -0.30 0.12 * -0.18 0.12 -0.04 0.02 0.09 0.13 -0.35 0.24 -0.05 0.08 ldist -0.31 0.12 * -0.85 0.14 ** -0.02 0.03 -0.22 0.17 0.00 0.25 0.17 0.10 distexcon 0.09 0.01 ** 0.07 0.01 ** 0.01 0.00 ** -0.01 0.01 0.13 0.02 ** 0.00 0.01 Perc -0.06 0.01 ** 0.07 0.01 ** -0.02 0.00 ** 0.17 0.02 ** -0.26 0.03 ** -0.10 0.01 ** perc2 0.00 0.00 ** 0.01 0.00 ** 0.00 0.00 -0.01 0.00 ** 0.02 0.00 ** 0.01 0.00 ** tradcost -0.21 0.02 ** -0.06 0.02 ** -0.02 0.00 ** 0.06 0.02 ** -0.14 0.05 ** -0.05 0.01 ** lang -0.07 0.10 -0.31 0.10 ** 0.14 0.12 -0.09 0.07 landlocked -0.74 0.14 ** -0.03 0.14 ** -0.23 0.14 0.05 0.08 ExDays 0.04 0.01 ** 0.00 0.01 0.00 0.00 0.02 0.01 ExDocs -0.15 0.03 ** -0.16 0.03 ** 0.14 0.03 ** -0.09 0.02 ** Trade 0.14 0.01 ** 0.28 0.02 ** 0.01 0.00 -0.03 0.03 0.19 0.03 ** 0.02 0.02 Y08 0.18 0.11 0.02 0.12 0.00 0.02 0.06 0.21 Y09 -0.20 0.11 -0.24 0.12 * -0.02 0.02 -0.28 0.21 Y10 -0.10 0.11 -0.09 0.12 -0.02 0.02 -0.26 0.21 Y11 -0.10 0.11 0.03 0.12 -0.02 0.02 -0.20 0.21 Obs 7,436 17,657 376 2,997 376 2,997 R2 0.2348 0.2614 Adj R2 0.2329 0.2606 AIC 11,974 11,894 56 Table 2.9: Cotton Apparel Repeated Heckman Results 2007 2008 2009 2010 2011 Parameter Estimate SE Estimate SE Estimate SE Estimate SE Estimate SE Intercept 3.08 6.15 10.70 5.75 21.80 6.55 ** -6.66 6.25 11.93 5.91 * lgdp 0.71 0.20 ** 0.05 0.17 -0.08 0.19 1.03 0.18 ** 0.30 0.17 lgdpop -1.04 0.29 ** -0.52 0.22 * -0.37 0.26 -1.13 0.31 ** -0.78 0.25 ** LPI -0.21 0.66 0.12 0.51 -0.25 0.52 -0.57 0.86 -0.35 0.45 ldist -0.38 0.55 0.38 0.56 -0.66 0.60 -0.07 0.62 -0.09 0.57 distexcon 0.14 0.03 ** 0.12 0.04 ** 0.14 0.04 ** 0.16 0.04 ** 0.12 0.04 ** Perc -0.33 0.06 ** -0.27 0.06 ** -0.21 0.06 ** -0.44 0.06 ** -0.25 0.06 ** perc2 0.03 0.00 ** 0.02 0.00 ** 0.02 0.00 ** 0.03 0.00 ** 0.02 0.00 ** tradcost -0.23 0.10 * -0.16 0.09 -0.11 0.11 -0.22 0.12 -0.10 0.12 ExDays 0.01 0.03 0.03 0.03 0.05 0.03 0.01 0.03 0.00 0.03 Trade 0.16 0.07 * 0.16 0.07 * 0.17 0.07 * 0.18 0.07 * 0.21 0.06 ** Select Intercept -3.98 1.97 * -6.02 2.04 ** -10.88 2.07 ** -5.57 1.88 ** -6.77 1.96 ** Select lgdp 0.38 0.06 ** 0.33 0.06 ** 0.37 0.06 ** 0.35 0.06 ** 0.30 0.05 ** Select lgdpop -0.36 0.09 ** -0.24 0.07 ** -0.27 0.08 ** -0.36 0.09 ** -0.38 0.08 ** Select LPI 0.02 0.24 -0.25 0.17 -0.31 0.17 0.13 0.27 0.15 0.16 Select ldist -0.25 0.21 0.17 0.22 0.69 0.23 ** -0.05 0.21 0.35 0.23 Select distexcon 0.02 0.01 0.00 0.01 -0.02 0.02 0.01 0.01 -0.02 0.01 Select Perc -0.08 0.02 ** -0.11 0.03 ** -0.13 0.03 ** -0.10 0.02 ** -0.11 0.03 ** Select perc2 0.01 0.00 ** 0.01 0.00 ** 0.01 0.00 ** 0.01 0.00 ** 0.01 0.00 ** Select tradcost -0.06 0.04 -0.07 0.03 * -0.10 0.03 ** -0.03 0.03 -0.01 0.03 Select lang -0.15 0.16 -0.13 0.14 -0.24 0.16 0.02 0.14 -0.01 0.15 Select landlocked 0.40 0.19 * 0.06 0.18 -0.28 0.18 0.27 0.17 -0.10 0.18 Select ExDocs -0.04 0.05 -0.08 0.04 -0.10 0.05 * 0.00 0.04 -0.13 0.05 ** Select Trade -0.03 0.03 0.05 0.04 0.09 0.04 * -0.01 0.03 0.02 0.03 AIC 2,450 2,459 2,358 2,366 2,397 N (OLS / Select) 55 / 596 81 / 609 77 / 609 55/ 588 99 / 595 57 Table 2.10: Animal Apparel Repeated Heckman Results Parameter 2007 SE 2008 SE 2009 SE 2010 SE 2011 SE Intercept -15.67 5.62 ** -14.27 5.62 * -16.90 5.40 ** -14.89 5.30 ** -16.83 5.21 ** lgdp 0.64 0.15 ** 0.50 0.16 ** 0.74 0.15 ** 0.47 0.14 ** 0.65 0.13 ** lgdpop -0.21 0.27 0.00 0.18 -0.02 0.22 -0.28 0.25 -0.05 0.21 LPI 0.52 0.70 0.03 0.45 0.12 0.50 0.75 0.75 0.17 0.46 ldist 0.22 0.54 0.65 0.51 -0.04 0.55 0.65 0.54 0.27 0.53 distexcon 0.00 0.03 0.00 0.03 0.00 0.04 -0.07 0.04 * -0.04 0.03 Perc 0.03 0.07 0.11 0.06 0.09 0.07 0.19 0.07 ** 0.06 0.07 perc2 0.02 0.00 ** 0.01 0.00 * 0.01 0.00 ** 0.01 0.00 ** 0.02 0.00 ** Unit 0.00 0.03 -0.03 0.03 -0.03 0.03 -0.05 0.03 * 0.05 0.03 unit2 0.00 0.00 . 0.00 0.00 0.00 0.00 . 0.00 0.00 . 0.00 0.00 . tradcost -0.02 0.10 -0.01 0.08 0.15 0.10 0.11 0.09 0.23 0.07 ** ExDays 0.01 0.02 0.04 0.02 0.05 0.02 * -0.02 0.02 0.03 0.02 Trade 0.35 0.07 ** 0.31 0.06 ** 0.30 0.06 ** 0.40 0.06 ** 0.32 0.06 ** Select Intercept -7.02 1.33 ** -6.79 1.44 ** -5.70 1.35 ** -6.71 1.32 ** -5.87 1.34 ** Select lgdp 0.20 0.04 ** 0.24 0.04 ** 0.24 0.04 ** 0.17 0.04 ** 0.24 0.03 ** Select lgdpop 0.03 0.06 -0.06 0.05 -0.06 0.05 -0.03 0.06 -0.04 0.05 Select LPI 0.04 0.17 -0.10 0.13 -0.07 0.12 0.09 0.19 0.01 0.12 Select ldist -0.01 0.14 0.06 0.16 -0.08 0.14 0.15 0.14 -0.12 0.14 Select distexcon 0.00 0.01 0.00 0.01 0.00 0.01 -0.02 0.01 * 0.00 0.01 Select Perc 0.04 0.02 * 0.02 0.02 0.02 0.02 0.04 0.02 * 0.02 0.02 Select Perc2 0.00 0.00 ** 0.00 0.00 ** 0.00 0.00 ** 0.00 0.00 ** 0.00 0.00 ** Select Unit 0.00 0.01 0.00 0.01 0.00 0.01 -0.02 0.01 * 0.01 0.01 Select Unit2 0.00 0.00 . 0.00 0.00 0.00 0.00 . 0.00 0.00 . 0.00 0.00 Select tradcost 0.00 0.03 0.00 0.02 0.03 0.02 0.06 0.02 * 0.06 0.02 ** Select lang 0.05 0.07 -0.06 0.10 -0.14 0.08 0.00 0.08 0.12 0.07 Select landlocked 0.05 0.09 0.23 0.13 0.26 0.11 * 0.07 0.09 0.17 0.09 Select ExDocs 0.05 0.02 * -0.02 0.03 -0.03 0.03 -0.02 0.02 -0.02 0.02 Select Trade 0.08 0.02 ** 0.08 0.02 ** 0.09 0.02 ** 0.11 0.02 ** 0.05 0.02 ** N (OLS / Probit) 112 1190 62 1165 74 1165 22 1127 107 1140 AIC 3184 3134 3162 3151 3210 58 Table 2.11: Cotton Apparel Market by Source Country 2007-2011 Country Custom Value Share China $6,324,420,108 23.4% Honduras $2,877,968,809 10.6% Mexico $2,305,943,515 8.5% El Salvador $2,026,551,389 7.5% Vietnam $1,596,953,983 5.9% Pakistan $1,355,572,934 5.0% Indonesia $1,188,499,148 4.4% India $1,122,014,956 4.2% Haiti $974,719,506 3.6% Cambodia $882,247,968 3.3% Table 2.12: Marginal Effect of Perc In Cotton Apparel Market Country Perc Marg. Eff. China 11.16 -0.0073 Haiti 0 -0.0375 Honduras 0 -0.0255 Table 2.13: Export by Apparel Type China Honduras Haiti Coat $3,281,095,069 $78,260,905 $68,527 Suit $10,985,292 $0 $0 Sweater $1,362,239,645 $42,029 $19,116 Tshirt $1,670,100,102 $2,799,665,875 $974,631,863 Total $6,324,420,108 $2,877,968,809 $974,719,506 59 Table 2.14: Cotton Coats, Sweaters and T-Shirts Marginal Effects 2007-2011 Coats Sweaters T-Shirts Marginal Effect Marginal Effect Marginal Effects Coeff. China Haiti Coeff China Haiti Coeff China Haiti Perc -0.19 ** -0.35 ** -0.30 ** Select Perc -0.09 ** -0.000 -0.034 0.01 -0.000 -0.004 -0.09 ** -0.031 -0.006 distexcon 0.16 ** 0.06 ** 0.18 ** Select distexcon 0.07 ** 0.000 0.028 -0.01 0.000 0.003 0.00 0.001 0.000 OLS Obs. 116 114 57 Probit Obs. 857 856 428 AIC 3487 3487 1968 Marginal Effect taken from average value respective country?s variable 60 Figure 2.1: Panel A: Homoskedasticity check for Level OLS model without observations of zero Panel B: Homoskedasticity check for Logged OLS model with observations of zero 61 Chapter 3 The Work Opportunity Tax Credit 3.1 Introduction In January 2013, Walmart President and CEO Bill Simon announced a focused initiative to hire any U.S. veteran who is honorably discharged within the first 12 months of his or her separation from active duty. Walmart projects it will hire over 100,000 veterans over the next five years. The initiative is set to begin Memorial Day of 2013 (Walmart, 2013). This announcement came less than two weeks after President Obama confirmed the latest extension of the Work Opportunity Tax Credit (WOTC) program when he signed the American Tax Payer Relief Act of 2012 (ATPRA). The WOTC is a tax credit to encourage employers to hire from targeted groups of people. Veterans are one of those groups. The WOTC was last extended through The Vow to Hire Heros Act of 2011 (VOW), but only for veteran groups. ATPRA brings back eligibility for non-veteran groups. Several business and labor leaders have praised Walmart?s commitment, while others have taken a more cynical approach toward the hiring pledge (Merrick, 2013). Commentators have projected these hires would have occurred anyway and Walmart is trying to gain social favor while receiving a windfall of subsidized labor (Matthews, 2013). A key part of Walmart?s pledge is the commitment to hire ?any? honorably discharged veteran, which would accordingly include disabled veterans. The largest tax credit offered by the WOTC legislation is targeted toward those with service-connected disabilities. This group, disabled veterans, is thus affected by multiple pieces of legislation and programs. Veterans with service-related disabilities are eligible for the U.S. Department of Veteran Affairs? Disability Compensation. The compensation varies at 10% increments according to the 62 determined percentage disability of the veteran and the number of the veteran?s dependents. The compensation ranges from $129 (10% disability, no dependents) to over $1000 per month for a veteran with dependents and 50% determined disability12. Autor and Duggan (2007) found an extension of benefits through this program to Vietnam Veterans in 2001 triggered early retirements, or exit from the labor force due to a newly-created, increased reservation wage. Similarly, Autor (2011) found disabled non-veterans have less incentive to enter the work force due to benefits from the U.S. Social Security Disability Insurance (SSDI) program, which he claims provides an average benefit of over $1000 per month. The other legislation for consideration is the Americans with Disabilities Act of 1990 (ADA90) and its amendments from 2008 (ADA08). The ADA90 and ADA08 give anti-discrimination protection to disabled Americans and require employers to make ?reasonable accommodations? for disabled employees. Previous studies suggest employers may avoid hiring the disabled in order to avoid accommodation costs. The WOTC is the subject of several studies and reports. In response to a U.S. Congressional request, the General Accounting Office13 (GAO) conducted a survey (Wozny & Daly, 2001) of Texas and California companies to determine if employers intentionally replace employees to take advantage of the WOTC. The report concluded less than 10% of employers found more than a little cost effectiveness in of ?dismissing? or ?churning? employees to take advantage of the WOTC. However, the same report also found 20% of companies offered incentives to managers to hire WOTC eligible employees. Hamersma (2003) largely confirmed the GAO 12 The benefit goes up to and beyond $3000 per month for those with 100% disability. 13 Later changed to Government Accounting Office 63 report in a national study of the WOTC and the Welfare to Work Tax Credit14 (WtWTC) concurrently using 1997 and 1999 data, finding low participation rates of employees in both credits. Gunderson and Hotchkiss (2007) performed a study to determine differences of employment separation time between WOTC eligible employees and non-WOTC eligible employees. The study is unique in that it comes from a single, large employer?s data and is narrowed to those working part-time in the same function between 1998 and 1999, two complete years. They found WOTC-eligible workers tend to stay slightly longer than non-WOTC workers, albeit by only a matter of days. The authors broke down reasons of separation by cause between the two classes of workers. An important finding of the report from summary statistics comes from the reason for separation. The average tenure of employment was 47 days15 for all subject employees, 46 days among WOTC-eligible employees. The standard deviation for each reason of separation ranged from 36 to 59 days. Simply using the average separation time and a 250 day work year, results in an opportunity for the employer to hire a WOTC eligible employee five times per year, provided one applies for the position. Keeping in mind these were part-time employees working 4-5 hours per day, many times the employer was not able to meet the 120 hour minimum to receive any of the WOTC benefit. Heaton (2012) found disabled veterans gained a roughly 2% increase in employment from the VOW legislation using a triple- and quadruple- difference approach, or DDD and DDDD. This increase in employment percentage translated to an estimated 32,000 jobs for disabled veterans. Those with cognitive disabilities were found to have benefits from the WOTC in DDD results, but no statistically significant change with the DDDD findings. Those with non- 14 The WtWTC was eventually rolled into the WOTC 15 Weighted-average of averages from Table 2, page 325 of Gunderson and Hotchkiss (2007) 64 cognitive disabilities saw near identical increases with both approaches. Further, Heaton found among disabled veterans, those receiving Social Security Disability Insurance (SSDI) benefits had no increase in employment, while those not receiving Social Security payments saw an increase of nearly 4% employment. Finally, the author found the employment increases of disabled veterans were in the full-time category rather than part-time. The present study uses a DDD approach to estimate if the WOTC is simply a windfall profit opportunity for employers, such as Walmart, or if it assists veterans, in particular disabled- veterans, to gain income through increased employment, wages or both. The comparison will use empirical analysis based on the changes to the WOTC from VOW to estimate the expected changes in employment from the WOTC extension through ATPRA. The empirical analysis is done with the data in which the veteran and non-veteran have WOTC eligibility, 2010, and then non-veterans lose credit eligibility and veterans maintains credit eligibility, 2011. For clarification, this period will not show up with cursory view of future IRS records because employers are able to retroactively claim the credit for the brief time period in which it was not available for non-veterans hired who eventually became eligible. After passage of the ATPRA, much of the perceived advantage for the veteran from VOW is removed. This should give insight to how much of an advantage the veteran is now losing with the passage of ATPRA. 3.2 Legislative History The Work Opportunity Tax Credit (WOTC) was originally authorized for one year, between October 1st, 1996 and September 30th, 1997 as part of the Small Business Job Protection Act of 1996. The WOTC targeted seven groups of the unemployed for assistance through tax credits to employers who hire them. The program was revised and extended eight times between 1997 and 65 2009 to eventually include 12 target groups through the American Recovery and Reinvestment Act of 2009 (ARRA), including two new group of veterans who are neither disabled from service nor receiving SNAP benefits qualify for a maximum credit $2400 or $5600 depending on their duration of unemployment. VOW maintained a maximum credit of $2400 for veterans receiving SNAP16 benefits. Veterans with a service-connected disability (DVET) qualified for a $4800 maximum credit to the employer, but VOW increased this amount to $9600 if the veteran was unemployed for 6 months or more. Finally, tax-exempt nonprofit (503(c)) organizations became eligible to receive the WOTC hiring credit when hiring qualified veterans. All these provisions were made effective for veterans hired November 22, 2011 through December 31, 2012 (Labor, 2012). ATPRA retroactively brought back the WOTC for targeted non-veteran groups, leaving both veterans and non-veterans eligible. A summary of this is in Table 3.1. The WOTC is provided to employers as a percentage of wages paid to the qualified employee and is based on the time the employee is employed. For example, 25% of the first $24,000 worth of wages paid to a newly hired disabled veteran (DVET) who has been unemployed for 6 months is qualified for a federal credit if the DVET works at least 120 hours during the first year of employment. The percentage increases to 40% if the DVET works a minimum 400 hours. Companies are encouraged to hire WOTC-eligible employees early in the tax year in order to maximize the amount of qualified wages. 16 Supplemental Nutrition Assistance Program 66 3.3 Theoretical Framework Allowing i to represent the labor group, the competitive model put forth by Acemoglu and Angrist (2001) can be altered to consider the effects of hiring credits on disabled veterans from the WOTC under the VOW legislation. Equation 3.1) is the modified version of the Acemoglu and Angrist model: 3.1) Max ? (D/Vt, N/Vt) = ? ?t(F(N/Vt, eD/Vt,) (a) Output ? wd,tDt ? wdv,tDVt ? wnv,tNVt ? wn,tNt (b) Wages ? aDt ? aDVt + cDVt + cVt (c) Accommodation costs and credits ? fdsDt-1 ? fdvsDVt-1 ? fnvsNVt-1 ? fnsNt-1 (d) Firing costs ? hd{DF ? [Dt ? (1-s)Dt-1]} (e) Disabled applicant hiring costs ? hdv{DVF ? [DVt ? (1-s)DVt-1]} (f) Disabled Veteran applicant hiring costs ? hnv{NVF ? [NVt ? (1-s)NVt-1]} (g) NonDisabled Veteran applicant hiring costs - hn{NF ? [Nt ? (1-s)Nt-1]}) (h) NonDisabled applicant hiring costs Acemoglu and Angrist (A&A) suggest firms need to replace non-productive workers with probability s each period t. Hiring costs consist of the probability of being sued pi at the cost of vi for not hiring an applicant who eventually sues for hiring discrimination. Hiring costs are thus hi = pi*vi, the cost of not hiring applicant iF. Firing costs are similarly developed from the probability of being sued for wrongful termination, fi = qi*?i. In this case qi and ?i are probability and cost, respectively, of a law suit for wrongful termination of the s non-productive employees. Workers increasingly supply labor li according to the function li(wi), with wi the wage received to the worker. Nondisabled workers N are assumed to represent 100% capacity of the individual labor function while disabled workers (D or DV) produce at some level e relative to N (or NV), so that a disabled employee is never more productive than a nondisabled employee, or where e ? 1 and eD/V ? N. Disabled or disabled veteran employees? marginal productivities can be increased amount B at cost A, up to 100% that of the nondisabled. The cost of the (reasonable) accommodation is voluntarily acceptable to the employer as long as B > A. Net 67 accommodation cost to the firm is a= A - B. The extension of the A&A model comes from consideration of the WOTC credit for disabled veterans, cDVt and also for the nondisabled veterans, cNVt. Finally, firms will enter the labor market provided profit ?, which is discounted at rate ?, exceed entry costs ?. This situation sets up a two-period17 profit maximization for the firm, t and t+1. For this study, t = 2011, t+1=2012. The difference between these two periods should give insight to period t+2, 2013. The cost of a law suit for not hiring an applicant is automatically avoided when the applicant is hired by the employer. Also, since the firm is replacing non-productive employees s from period t-1 with productive employees in period t, hiring costs are considered subsidies and increase profit, or fi < wi, the condition which triggers employee replacement. Of course growth causes hiring, but for the purposes of this analysis, the hiring process is indifferent to the causes of hiring. Setting the first order conditions from equation 3.1) to zero for each source of labor sets up an equilibrium condition to measure the hiring decision of the firm: 3.1?) FOCs: FD = wd + ?sfd ? [1-?(1 ? s)]hd + a (a) Disabled market clearing condition FDV = wdv + ?sfdv ? [1-?(1 ? s)]hdv + a - c (b) Disabled Veteran market clearing condition FNV = wnv + ?sfnv ? [1-?(1 ? s)]hnv ? c (c) NonDisabled Veteran market clearing condition FN = wn + ?sfn ? [1-?(1 ? s)]hn (d) NonDisabled market clearing condition The General Rule of the ADA90 and ADA08 (Justice, 2009) requires employers to compensate employees equally without regard to disability status and provide reasonable accommodation which would permit the disabled (or disabled veteran) employee?s productivity to equal that of the nondisabled. Using lines (a) and (d) in Equation 3.1?), this would require wn = wd. The ADA90 (and ADA08) provide hiring and firing protection for disabled, which requires hd(?) 17 Acemoglu and Angrist present an indefinite period model. 68 fd(?) in an amount equal to the cost of accommodation a. Empirical results from nationally performed studies (DeLeire 2000; A&A 2001) suggest employers avoid the accommodation costs because the hiring and firing costs are not realistically high enough for employers to hire the disabled due to ADA90. In essence, many empirical results suggest a > wd + ?sfd ? [1-?(1 ? s)]hd or, the cost to not hire a disabled worker is less than the cost of accommodation. The WOTC presents another consideration though. Line (b) and line (c) of equation 3.1?) include a credit for employers, c. This sets up a situation for comparison of four groups of employees for employer reaction to accommodation costs and hiring credits. The impact of the WOTC depends on how employers react to the wage subsidy in the short- term, as the credit applies to wages mostly during the 2012 calendar year. The credit is assumed as an attempt to shift the demand curve outward, since it is directed at the employer rather than the applicant as a wage subsidy, which would shift the labor supply curve out. A measure in the change of wages and labor hours should give insight into which side of the labor market gains more of the incidence from the tax credit, employers, through more labor at the same or reduced paid wage, or employees through increases in wages. Since the WOTC is not a permanent credit, employers would likely be hesitant to increase wages. However, a counter-argument to employer hesitation to increase wages is the continual renewal of the WOTC every 1-3 years. As noted in the introduction, previous studies indicate employers avoid hiring the disabled in order to avoid accommodation costs. DeLeire, Acemoglu and Angrist and Jolls and Prescott (2004) found negative consequences for employment of the disabled due to new legal considerations for accommodation costs to employers. However, the WOTC brings with it a 69 credit for hiring veterans with service-related disabilities18, creating four sources of labor for employers to consider: 1) Disabled (D) 2) Disabled Veterans (DV) 3) NonDisabled Veterans (NV) 4) NonDisabled (N) Group 1 consists of members of the labor force who are disabled and without veteran status. Group 2 consists of members of the labor force who are disabled and with veteran status. Group 3 consists of members of the labor force who are not disabled and have veteran status. Group 4 consists of members of the labor force who are not disabled and are not veterans. No individual is a member of more than one group. 3.4 Data and Empirical Analysis Data for analysis come from the Current Population Survey (CPS) available from the U.S. Census Bureau (2012). The CPS is a monthly survey consisting of several individual and household characteristics for income and demographics. The unit of observation is an individual between the ages of 25 and 55 years-old (inclusive). The WOTC legislation renewal became effective in late November 2011. The data set includes the five months from December 2011 through April 2012 while the increased hiring credits were available for hiring veterans, and newly unavailable for hiring non-veterans. Unemployment dropped from 8.5% to 8.1% during this time period. Accordingly, the data set also includes the five-month period December 2010 18 The WOTC extension also does apply to veterans based on their participation in the SNAP who are not disabled, but they are not identified differently from other veterans. 70 through April 2011 as the time period prior to the hiring credit increases, when the unemployment rate dropped from 9.3% to 9.0%. The VOW legislation was not approved until late 2011, so employers operated through most of 2011 without confirmed knowledge of the increases to credits for hiring veterans over non- veterans. Likewise, the ATPRA was not signed into law until January 2013, retroactive effect back into 2012. This indicates employers would be operating with similar knowledge of the WOTC extensions in both time periods. Four outcomes are used as indicators of how employers react to the increased hiring credit, two for earnings and two for employment. Equation 3.2) represents the full empirical model for estimation: 3.2) yit = ?0 + ?1Git + ?2Eit + ?3WOTCt + ?4DitWOTCt + ?5EitWOTCt + ?6Xi G represents the individual?s group status through a categorical variable for group status. In line with the theoretical model, individuals are segmented as veterans VET, nondisabled NON, disabled DIS or disabled veterans DVET. E represents a categorical variable indicating the individual?s highest level of completed education with levels of high school or less, an associate?s degree, a bachelor?s degree and a graduate degree. The impact of unobserved factors between the two time periods is captured by the WOTC variable, set to 0 for the first time period and 1 for the time after the VOW was implemented. This is then interacted with the group and education variables to estimate the effects of the increased hiring credits. Finally, X represents a set of demographic variables for sex and race, along with a categorical variable for job sector as additional controls. Table 3.2 has demographic explanatory variables. AGE is normalized to 1 for a 25 year-old and 31 for a 55 71 year-old. This is done to avoid a negative intercept for earnings results and make interpretations more straight forward. The CPS gives survey respondents seven options indicating employment status: 1 ? Employed, At Work 2 ? Employed, Absent 3 ? Unemployed, On Layoff 4 ? Unemployed, Looking 5 ? Not In Labor Force, Retired 6 ? Not In Labor Force, Disability 7 ? Not In Labor Force, Other For the purposes of this study, option 6 is considered unemployed and grouped with option 3 and option 4. Those who respond to the survey with option 5 or option 7 are not considered in the data set. Also, those who respond as unemployed are considered to work 0 hours per week for the UslHrs dependent variable. Those who respond to work less than 20 hours per week are considered unemployed for the employment outcome. The result is a larger labor force than used by the U.S. Bureau of Labor Statistics, and thus a higher unemployment rate than it would report. Table 3.3 includes an educational summary for each group. Table 3.4 includes four panels of earnings and employment summary statistics for each group. The earnings statistics reflect only those observations in which the respondent was employed. 3.5 Regression Estimates The impact of the VOW legislation is measured with a DDD approach. The differences of primary interest are group membership and educational attainment. Veterans have an educational bias due to enlistment requirements for the military. Segmenting by educational attainment provides for determination of employment increase due to veteran status, or because 72 of the education bias. An important assumption of the data is that employers hired in 2012 with the understanding the WOTC was only extended for veterans. Four dependent variables are estimated to gain insight on how employers react to the hiring credits for veterans, with or without disabilities. Hourly and weekly wages are estimated as earnings rates. Employment is first measured according to a binary distribution of employed or unemployed. Then employment is estimated with a categorical approach where usual hours worked is broken into four categories. Categorical variables for group membership (Group) and educational attainment level (Educ) are estimated separately and interacted with each other. The WOTC term identifies the change in the relative dependent variable between time periods, the period December 2010 through April 2011 (before WOTC changes) to the period December 2011 through April 2012 (after WOTC changes). Age, sex, race and geographic location are also controlled for in the estimating models. The wage models also consider the industrial sector of employment. Age is the only continuous variable used in the estimating functions. The base individual for comparison is a 25 year-old non-white non-disabled female non-veteran, with a high school diploma or less, living in the South. This person is unemployed if estimating one of the employment measures, and this person is working the Trade industry if estimating wages. Due to the size of the data set, statistical significance is considered at 5%. The hourly wages appear to follow close to a lognormal distribution, while weekly wages appear closer to a gamma distribution19. Both dependent variables were regressed under multiple distribution assumptions, including a negative binomial. Table 3.5 displays model fit criteria results for each distribution assumption. Based on the AIC comparisons, the gamma distribution is used for hourly wages and the negative binomial distribution is used for weekly wages. 19 Distribution graphs are displayed in Appendix C 73 The Usual Hour distribution is centered around 40 hours, as would be expected, but it does not have a well-defined distribution. Therefore, employment is estimated as binary, employed or not employed, and then with four categories of employment status: Unemployed: < 20 hours per week Underemployed: 20 ? 31 hours per week Employed: 32 ? 50 hours per week Overemployed: 51+ hours per week The more detailed analysis for employment status is done with a baseline categorical logit model, using ?Unemployed? as the base. Results are broken into the period before WOTC and the period after WOTC. A baseline logit model can be somewhat cumbersome for interpretation. The baseline is the base individual?s log odds of the presented status versus being unemployed. The base individual thus has three intercepts, one for comparison of unemployment as a base to each other employment status category. Interpretation of the coefficient is done by taking its antilog. The result is the odds of the employment status category versus the base category of unemployment: where ?j is the intercept for the considered status, ?j is a vector of coefficients for the comparative category, and Xi is a vector of the ith individual?s characteristics. For example, an intercept for employed may be -0.09. This would be interpreted as the baseline individual?s probability of being employed is 91% of being unemployed. Adding any education makes the base individual much more probable of being employed versus unemployed. 74 3.5.1 Regression Results Table 3.620 summarizes the regression results for wages and employment categories using statistically significant coefficients from the regression estimates. The weekly wages have more reporting observations, and are thus more robust than the hourly wages. However, the hourly wages do provide an indication of whether or not the credit passes through to the employee through earnings rate, or employment rate. In general, employment and income increase with education. Also, the disabled earn less than the non-disabled. An indication of disabled veterans with an associate?s degree gaining in from the WOTC can be found in weekly wages. However, disabled non-veterans with the same education level apparently lost income. This may indicate disabled veterans gained from the WOTC from pulling employment away from disabled non- veterans. Moving left to right in the employment sections of Table 3.6, the biggest increases of employment came from the veteran group, within the underemployed category displayed in Panel D. This would indicate some veterans moved from unemployed to underemployed. That movement may be an example of the marginal benefit to employers the WOTC was intended to provide in order to increase hiring. A look within the employed category may provide some insight on employer reactions to the WOTC between the two disabled groups. The middle rows of Panel E contain the differences of employed compared to unemployed. The disabled veteran becomes more probable to be employed while the disabled non-veteran becomes less probable to be employed. This could be an indication of employer preference changes due to the WOTC changes. Disabled veterans saw employment increase, but at the cost of disabled non-veterans at each education level. Panel E of Table 3.6 also indicates the preference for veterans diminishes as education increases. Non- 20 Coefficient estimates for the full models? results are available in Appendix C 75 veterans have lower probabilities of being employed at the lowest education level. This condition does not hold as education increases though, likely a condition from the previously noted upward education bias for veterans due to military entrance requirements. In general, Table 3.6 suggests the WOTC changes may have helped increase employment of veterans, without a decrease of employment to non-disabled non-veterans. Disabled veterans gained employment after the WOTC changes, but these gains may have come at the expense of the non-veteran disabled. 3.6 Regression Estimates for Trade Industry Sector Section 3.5 provides some insight on how employers reacted to the extension of the WOTC from the Vow to Hire Heros Act of 2011. In order to evaluate expectations of the most recent extension of the WOTC in regards to Walmart, the data set is narrowed to only those in the ?Trade? industrial sector, those employed in wholesale and retail trade. This changes somewhat how employment changes are evaluated. Section 3.5 included the entire dataset population. Respondents, who report to work any hours, also identify the industrial sector in which they work. This is what allows for segmentation of the ?Trade? industry. Of course, those who respond to working zero hours do not have an industrial sector to report, leading to an adjustment of the unemployment from less than 20 hours to greater than zero, but less than 20 hours21. Regression results are again collectively summarized for wages and employment categorical changes. A direct comparison to the overall economy results should be limited, because of the change to how ?Unemployed? is defined. However, similar overall hierarchal results for education and group membership are evident in regard to wages in the Trade industry as in the results from Section 3.5. 21 A summary of the changes of employment categories is available in Appendix C. 76 Table 3.8 displays summary regression estimate results for differences within and between the group membership and education levels for each time period. Again using weekly wages as the more robust measure, income remains constant for the most part between time periods for each group and education level. The most noticeable difference from Table 3.8 is the large increase in odds to be employed for disabled veterans. Again, the point estimates of employment are of secondary concern. The estimated employment difference is of primary interest for this study. There is also an indication of some employment increases to disabled veterans coming at the cost of less employment for the disabled non-veteran, from the binary employment results in Panel C. The categorical results in Panel E show a gain in employment for disabled-veteran, but seem inconclusive if it was at the cost of disabled non-veterans. The large increase in odds of full-time employment for a disabled veteran in the ?After? period suggests the WOTC changes perhaps provided the marginal benefit to employers needed to hire more disabled veterans. The increase in full-time employment instead of part-time is in line with the noted findings from Heaton. 3.6.1 Regression Analysis Speculatively, the increase in full-time employment for DVETs may also be the point at which disabled veterans become more willing to accept a job offer. Referring back to the Veterans Disability Compensation program noted in Section 3.1, a veteran without dependents and a determined 10% disability does not receive very much compensation on a monthly basis, and thus would only need a small increase in offered wage to accept an employer?s job offer. Likewise, the WOTC amount for disabled veterans does not distinguish between percentage 77 increments of disability, so a 10% disabled veteran garners the same credit for the employer as a 60% disabled veteran. From Equation 3.1), the necessary accommodation cost may be exceeded by the WOTC received by the employer. Prior to VOW, the WOTC credit difference to the employer for hiring a disabled veteran over non-disabled non-veteran target group member was $300 to $2400 in most cases over a one-year period. After VOW, the difference became at least $2400 and up to $9600, apparently creating a condition much more favorable to the disabled veteran in the Trade industry versus the general economy. The weekly wage results from Table 3.8 suggest stagnant wages in both periods of the study for disabled veterans. As a result, the employer should expect to capture the vast majority of the benefit from the WOTC, with only a small marginal increase in wage to the disabled veteran, but just enough to surpass the reservation wage and enter the employed category. This concept is depicted in Figure 3.1, where the employer demand curve moves out from equilibrium with an inelastic disabled veteran labor supply curve, beyond the reservation wage, and to an elastic supply curve. Figure 3.1: Change in Wage Received and Wage Paid for Disabled Veterans 78 The distance between Wage0and WageEMP represents what would be considered windfall profit to the employer. However, since the WOTC is only paid out once per hire, the windfall profit would be reduced each year by the difference between WageDVET and Wage0. In order for the employer to maintain the separation between Wage0 and WageEMP, a new employee would need to be hired each year. Or, the employer could invest in the DVET to reduce exposure to labor turnover costs by keeping the wage to the DVET above the reservation wage. 3.7 Conclusions and Analysis The Work Opportunity Tax Credit was extended through the American Tax Payer Relief Act of 2012, continuing a credit to employers who hire targeted veterans within their first five years of discharge from active service. The ATPRA also brings back the WOTC for targeted non- veteran groups. This study uses a difference-in-different approach with data from extension of the WOTC for veterans through the Vow to Hire Heros Act of 2011, but also removal for non- veteran groups, to estimate the impact this will have on earnings and employment of veterans and non-veterans according to disability status, creating four response groups. Each group is also evaluated with its educational attainment level to better narrow down those who will be affected by the WOTC extension. Additionally, estimations are narrowed to the retail and wholesale trade industry (Trade) and evaluated to estimate the effect Walmart will have on its announced efforts to hire recently discharged veterans. Estimation results show an inherent hierarchy of earnings and employment according to veteran status, disability status and education level, as expected. For those with a high school education or less, the disabled earn less and are less-employed than the non-disabled. And, veterans earn and work more than their non-veteran counterparts. These tendencies diminish as 79 education increases. However, the regression results show clear evidence of two different labor market equilibriums, one for the non-disabled and one for the disabled, at each level of education. In the general economy, increases for non-disabled veterans in the underemployment level between 20 and 31 hours per week, with growth in weekly income rates outpacing those of non- veterans occurred when the WOTC became eligible only for veterans. Disabled-veterans at the associate?s degree education level also increased income and employment as a result of WOTC changes, but some of these gains may have come from disabled non-veterans. Some potential of unintended consequences from VOW become apparent after review of regression results. The increases of employment of non-disabled veterans through the underemployed category may be an example of some games between veterans and employers, where a Nash equilibrium is reached. The WOTC is provided to employers based the number of hours the veteran works, with the maximum benefit reached at 400 hours. Veterans returning from service may need to establish civilian employment history, and earn a little income, while looking for a full-time job or attending college. If a windfall to the employer occurred, this is the situation where it could be most likely to have happened. At 40 hours per week, the 400 hour limit is reached in ten weeks with one employee. However, if the employer hires two veterans, each to work 20 hours per week (fitting the underemployed category), the employer has two opportunities hit the 400 hour limit within a 12 month period, and receive two credits. Since the WOTC would trigger hiring for employers who are at the margin of the hiring decision, reasonably only one full-time employee is needed. In this scenario, the first credit could take the employer to the point where marginal cost is equal to marginal benefit. The second credit would 80 then be windfall profit. Empirical evidence from this study and much of the existing literature indicate this scenario does not occur often, or at least not intentionally. Of course, if this was happening, it was probably already happening before VOW. Hamersma (2010) noted evidence employers may act differently to retain WOTC employees who near the 400 hour maximization point for the credit. However, she also notes employers do not seem to have much influence over maximizing the available credit because that is more influenced by the individual employee. Much of the literature regarding employee retention agrees with Hamersma and indicates the employee has more influence over the windfall to the employer by choosing to show up for work or not. The changes from VOW would have simply decreased whatever influence employers had to only those who hire veterans, thus creating a more favorable situation for veterans versus non-veterans. The passage of ATPRA brings back the credit for non-veterans, at an equal or lesser value. This should move the relative demand for veterans back toward that of the non-veteran. The Trade industry follows much the same initial patterns of the general economy for employment and earnings. Differences of wages for each evaluated group are estimated near zero, but disabled-veterans appear to have gained full-time employment after the WOTC changes. This finding is in line with Heaton (2012). The difference in this study is the evidence some of the gains may have come from employment of the disabled non-veteran, not through hourly wage increases. Nominally, the changes are small, so no noticeable changes in the labor market as a whole should be expected, but some changes in labor equilibrium should be expected for the two disabled groups. The apparent decrease in employment for disabled non-veterans should be considered a negative unintended consequence in the short-term, but may also have positive results in the 81 long-term. Again referring back to the profit maximization considered in Equation 3.1), the employer will not differentiate between disabled and non-disabled employees provided accommodations can bring the disabled up to the productivity levels of the non-disabled. In the Trade industry, the WOTC may provide enough coverage of accommodation costs to incentivize employers enough to make the accommodations for disabled-veterans. Since the WOTC is a one-time credit per hire, the benefit to the employer may not come through the credit itself, but from the capital asset used to accommodate the disabled-veteran and bring her up to par with the non-disabled. This accommodating asset, presumably, could then also be used by others who are disabled, whether veteran or not. This is depicted below where the disabled applicant has a similarly increased reservation wage due to the SSDI program as the disabled-veteran with the Disability Compensation program. Figure 3.2: Short-term and Long-term Potential Effects Both the disabled-veterans and the disabled can initially be on the same labor supply curve, S. Both the disabled and the disabled-veteran start atW0, where the equilibrium wage intersects with an inelastic labor supply curve, and employer demand D0. After VOW, the demand curve 82 for disabled-veterans may initially shift right relative to the disabled, and over the reservation wage, taking the DVET to an increased WDVET1 at an increased labor rate, LDVET1. The disabled, however, has the opposite reaction resulting at a lower wage and labor rate. This is the short-term same scenario depicted in Section 3.5. However, if the employer uses the WOTC to purchase an accommodating asset, then she will be able to offer an increased wage to disabled employees due to increased productivity. In other words, her demand curve will rotate up and become more elastic. Thus, the DVET will benefit from a shift out and a reorientation of the supply and demand elasticity relation, arriving at an increased equilibrium wage and labor amount. The shift out is the veteran effect. The reorientation is the accommodation effect. The shift out depends on the continued extension of the WOTC, but the accommodation effect would hold as long as the accommodating asset is viable. A disabled non-veteran should also see a benefit from the employer?s newly oriented demand curve as well. The return of the WOTC through the ATPRA moves the disabled back to the original demand curve of D0. The accommodating asset allows for the employer to accept an increased wage to the disabled because the accommodation is already in place from hiring the disabled-veteran. Both disabled groups arrive at a wage and labor rate above the reservation wage. For the individual employer, this may require the DVET to be hired, transition away from the company, and then be replaced by a DIS. However, an employer who runs multiple shifts would be able to employ both the DVET and DIS concurrently. Theoretically, according to Equation 3.1), the disabled applicant could actually compete with the non-disabled applicant and enter the higher paying non-disabled labor market. 83 Analysis of the general economy in this study indicates veterans may have benefited from the WOTC with increases in part-time employment while non-veterans were not eligible for the WOTC. Disabled veterans may have found benefit in the retail and wholesale trade industry during this same time, with full employment. The return of the WOTC through ATPRA for non- veteran groups would apparently reduce these relative benefits. Walmart?s commitment to hire veterans has drawn mixed reviews in the media, but its committed actions would apparently be beneficial to veterans beyond what itself would receive through the WOTC. If Walmart is representative of the ?Trade? industry, it was apparently already a particularly beneficial channel for disabled veterans and potentially disabled non-veterans, before its announced commitment. This does not mean Walmart is not set up for a windfall profit via the WOTC, in fact it apparently has opportunities beyond just hiring veterans. Walmart may have the opportunity to gain from disabled non-veterans who are indirectly more productive due to VOW and now re- eligible for the WOTC through ATPRA. It may be the windfall comes indirectly through increased productivity, rather than a direct tax credit. The WOTC has been in place to encourage employers to hire from targeted groups since 1996. Several studies have taken various angles of the WOTC. However, much of the literature eventually concludes with a speculation involving some sort of game between the employer and employee regarding the reservation wage and employee retention. Much of the literature shows employers do not necessarily react to the incentive, let alone benefit from it, because of the turnover from WOTC eligible employees. Windfalls from the WOTC seem more determined by legislation altering the reservation wage, such as unemployment insurance extension, than a direct consequence of WOTC changes. Referring back to Gunderson and Hotchkiss (2007), roughly one-third of WOTC eligible employees separated from the company within 20 days, or 84 within 100 hours, well under the 120 hours necessary for the employer to receive any benefit from the WOTC. Roughly 90% of the WOTC eligible employees terminated employment before the 400 hour maximization point22. The GAO survey (2001) reported average hiring costs to the employer for a WOTC eligible employee at $3800, and under $3300 for the non-eligible WOTC employee. Even if those costs to hire were halved, combined with the Gunderson and Hotchkiss findings, it would still likely be a questionable business practice for the employer to attempt to hire with much emphasis on the WOTC. This study?s contribution may be the speculative finding, with reasonable evidence to support profit maximization theory, employers will react to the more productive employee, not the temporarily lower wage. In order to make use of the WOTC funding, it may be a better policy to direct the funding to the employer in the form of a credit for work place accommodations, and to the employee in the form of a tax credit for remaining employed. 22 Refer to Figure 1, page 324 of Gunderson and Hotchkiss (2007). Hour estimations based on 5-hour workday. 85 Table 3.1: Work Opportunity Tax Credit Summary 2011 2012 2013 ARRA Maximum Tax Credits VOW Maximum Tax Credits ATPRA Maximum Tax Credits Target Groups & Wages 120 Hours Worked 400 Hours Worked 120 Hours Worked 400 Hours Worked 120 Hours Worked 400 Hours Worked Veteran Receiving SNAP $1,500 $2,400 $1,500 $2,400 $1,500 $2,400 Disabled Veteran hired within 1 year of discharge a $0 $0 $3,000 $4,800 $3,000 $4,800 Disabled Veteran unemployed 6 months $3,000 $4,800 $6,000 $9,600 $6,000 $9,600 Veteran unemployed 4 weeks $1,500 $2,400 $1,500 $2,400 $1,500 $2,400 Veteran unemployed 6 months a $0 $0 $3,500 $5,600 $3,500 $5,600 Short-Term TANF Recipient $1,500 $2,400 $0 $0 $1,500 $2,400 Long-Term TANF Recipient $5,625 $9,000 (over 2 yrs) $0 $0 $5,625 $9,000 (over 2 yrs) SNAP (food stamp) Recipient $1,500 $2,400 $0 $0 $1,500 $2,400 Designated Community Resident $1,500 $2,400 $0 $0 $1,500 $2,400 Vocational Rehabilitation Referral $1,500 $2,400 $0 $0 $1,500 $2,400 Ex-Felon $1,500 $2,400 $0 $0 $1,500 $2,400 SSI Recipient $1,500 $2,400 $0 $0 $1,500 $2,400 Veteran: ? A member of a family that received Supplemental Nutrition Assistance Program (SNAP) benefits (food stamps) for at least a 3-month period during the 15-month period ending on the hiring date, or ? Entitled to compensation for a service-connected disability hired within one year of discharge or release from active duty, or ? Entitled to compensation for a service-connected disability and unemployed for a period or periods totaling at least 6 months of the year ending on the hiring date, or ? Unemployed for at least 4 weeks (but less than 6 months) during the one year period ending on the hiring date, or ? Unemployed for at least 6 months during the one year period ending on the hiring date. Source: http://www.doleta.gov/business/incentives/opptax/benefits.cfm Source: http://www.doleta.gov/business/incentives/opptax/PDF/veterans_fact_sheet12_1_2011.pdf Source: http://www.irs.gov/Businesses/Small-Businesses-&-Self-Employed/ a Source: Public Law 112-56 (Vow to Hire Heros Act) 86 Table 3.2: Demographic Variable Summary Independent Variables Code Description Mean Std. Dev. Dem og ra ph ic AGE Age Years beyond the age of 24 Min: 1 = 25 year-old, Max: 31 = 55 year-old 16.5 9.022 RACE Race 1 = White only, 0 = Non White only 0.811 0.391 SEX Gender 1 = Male, 0 = Female 0.484 0.500 Indu str y Sect or INDUS Industrial 1 = Respondent employed in Agriculture, Forestry, Fishing, Hunting, Mining, Construction, Manufacturing, Transportation or Utilities, 0 = Otherwise 0.250 0.433 TRADE Trade 1 = Respondent employed in Wholesale or Retail Trade, 0 = Otherwise 0.126 0.332 INFO Information 1 = Respondent employed in Information services, 0 = Otherwise 0.023 0.151 PRFSNL Professional 1 = Respondent employed in Financial, Professional/Business services, Education or Health services, 0 = Otherwise 0.426 0.495 SRVCES Services 1 = Respondent employed in leisure, hospitality or other services, 0 = Otherwise 0.118 0.323 PUBADM Public Administration 1 = Respondent employed in Public Administration, 0 = Otherwise 0.056 0.230 Geo gr aphi c NRTH Northeast Region Maine, New Hampshire, Vermont, Massachusetts, Rhode Island, Connecticut, New York, New Jersey, Pennsylvania 0.199 0.400 MIDW MidWest Region Ohio, Indiana, Illinois, Michigan, Wisconsin, Minnesota, Iowa, Missouri, North Dakota, South Dakota, Nebraska, Kansas 0.232 0.422 SOTH Southern Region Deleware, Maryland, Washington D.C, Virginia, West Virginia, North Carolina, South Carolina, Georgia, Florida, Kentucky, Tennessee, Alabama, Mississippi, Arkansas, Louisiana, Oklahoma, Texas 0.315 0.465 WEST Western Region Montana, Idaho, Wyoming, Colorado, New Mexico, Arizona, Utah, Nevada, Washington, Oregon, California, Alaska, Hawaii 0.253 0.435 Employment and Sector Average and Standard Deviation based on those employed 87 Table 3.3: Educational Summary Group Associate Bachelors Graduate No Degree Total Disabled Veterans Frequency 459 301 132 2,187 3,079 Percent 0.1% 0.1% 0.0% 0.4% 0.6% Row Pct 14.9% 9.8% 4.3% 71.0% 0.0% Col Pct 0.8% 0.2% 0.2% 0.7% 0.0% Disabled Frequency 3,196 3,372 1,347 26,528 34,443 Percent 0.6% 0.6% 0.2% 4.8% 6.2% Row Pct 9.3% 9.8% 3.9% 77.0% 0.0% Col Pct 5.4% 2.7% 2.2% 8.5% 0.0% Non Disabled Frequency 52,133 114,798 57,275 266,047 490,253 Percent 9.4% 20.7% 10.3% 48.0% 88.4% Row Pct 10.6% 23.4% 11.7% 54.3% 0.0% Col Pct 87.4% 93.3% 93.8% 85.5% 0.0% Veteran Frequency 3,853 4,548 2,330 16,383 27,114 Percent 0.7% 0.8% 0.4% 3.0% 4.9% Row Pct 14.2% 16.8% 8.6% 60.4% 0.0% Col Pct 6.5% 3.7% 3.8% 5.3% 0.0% Total 59,641 123,019 61,084 311,145 554,889 10.8% 22.2% 11.0% 56.1% 100.0% 88 Table 3.4A: Disabled Veteran Employment and Earnings Educ Period Obs Variable N Mean Std Dev Median Min Max Associate After 256 HrWage 15 19.01 9.62 18.55 2.13 40 WkWage 19 934.26 514.22 740 316.66 2045 USLHRS 204 16.73 20.56 0 0 82 Before 203 HrWage 13 17.33 6.56 17.5 8.25 28.75 WkWage 20 765.93 423.05 681 247.5 1880 USLHRS 166 17.87 21.68 0 0 90 Bachelors After 145 HrWage 9 15.21 6.6 15 8.5 29.27 WkWage 18 1025.69 774.96 961.53 32 2884.61 USLHRS 119 23.15 19.78 30 0 80 Before 156 HrWage 7 26.64 16.5 24.5 9.72 60 WkWage 15 1400.76 885.95 1192 388.8 2884.61 USLHRS 115 21.58 23.22 15 0 90 Graduate After 69 HrWage 4 41.53 22.86 37 20 72.12 WkWage 10 1404.6 916.22 1204.61 106 2884.61 USLHRS 60 27.48 21.81 40 0 70 Before 63 HrWage 0 . . . . . WkWage 8 1612.11 821.37 1634.46 438.46 2884.61 USLHRS 57 25.58 22.52 40 0 60 No Degree After 1078 HrWage 48 16.3 6.73 15.75 7.25 36.65 WkWage 72 756.05 382.3 696 43.5 1865.38 USLHRS 895 12.67 19.87 0 0 99 Before 1109 HrWage 55 18.63 8.44 16 7 53 WkWage 63 812.41 556.8 694.4 8 2884.61 USLHRS 910 12.57 20.2 0 0 99 89 Table 3.4B: Disabled Employment and Earnings Educ Period Obs Variable N Mean Std Dev Median Min Max Associate After 1572 HrWage 105 15.88 7.65 13.97 3.01 45.75 WkWage 138 683.09 486.47 600 19 2884.61 USLHRS 1289 15.55 20.14 0 0 99 Before 1624 HrWage 114 16.47 9.16 13.68 6 51.15 WkWage 151 624.2 452.98 550 9.23 2307 USLHRS 1360 15.83 19.61 0 0 99 Bachelors After 1592 HrWage 85 19.78 12.92 15.75 2.35 72.12 WkWage 192 1003.73 675.57 820.26 40 2884.61 USLHRS 1258 21.09 21.41 18.5 0 91 Before 1780 HrWage 91 19.96 12.76 15.15 7 77 WkWage 195 982.8 702.81 843.75 0.23 2884.61 USLHRS 1441 21.94 21.11 25 0 99 Graduate After 681 HrWage 25 23.74 15.7 18 7.5 70 WkWage 89 1114.5 716.14 961.53 30 2884.61 USLHRS 563 26.03 20.95 35 0 80 Before 666 HrWage 26 27.93 15.63 25.88 7.5 69 WkWage 106 1223.08 722.89 1109.4 120 2884.61 USLHRS 558 27.49 21.23 40 0 80 No Degree After 13060 HrWage 567 13.2 6.59 11.25 1 65 WkWage 743 554.96 412.15 480 0.23 2884.61 USLHRS 10857 9.86 17.1 0 0 99 Before 13468 HrWage 594 13.06 5.95 11.35 2.1 52 WkWage 784 542.38 395 480 1 2884.61 USLHRS 11239 10.25 17.61 0 0 99 90 Table 3.4C: Non-Disabled Employment and Earnings Educ Period Obs Variable N Mean Std Dev Median Min Max Associate After 25899 HrWage 3274 17.91 8.43 15.78 2.11 72.12 WkWage 4949 807.17 476.22 708 0.03 2885 USLHRS 22108 36.35 13.92 40 0 99 Before 26234 HrWage 3484 17.89 8.75 15.8 2.13 96.15 WkWage 5114 801.89 493.03 692.3 5.76 2885 USLHRS 22330 36.34 13.58 40 0 99 Bachelors After 57532 HrWage 3964 20.8 12.12 17.5 1.01 99 WkWage 11075 1116.28 677.98 961.53 0.01 2885 USLHRS 48295 38.67 12.89 40 0 99 Before 57266 HrWage 3838 20.51 11.91 17.5 1.15 99 WkWage 10892 1104.49 668.19 961 0.01 2885 USLHRS 47936 38.48 13.3 40 0 99 Graduate After 28893 HrWage 1118 27.89 16.74 23.81 4.29 99.99 WkWage 5860 1406.96 742.1 1250 0.01 2885 USLHRS 25313 40.81 12.88 40 0 99 Before 28382 HrWage 1039 26.19 16.09 22 2.17 99.99 WkWage 5681 1372.54 730.62 1211.53 0.01 2885 USLHRS 24690 40.6 13 40 0 99 No Degree After 130600 HrWage 15627 14.72 6.94 13 1.01 98.85 WkWage 21835 680.3 442.71 576.92 0.01 2885 USLHRS 103034 33.98 16.46 40 0 99 Before 135447 HrWage 16230 14.6 7.08 13 1 99 WkWage 22545 656.31 424.83 560 0.01 2885 USLHRS 107023 33.47 16.58 40 0 99 91 Table 3.4D: Veteran Employment and Earnings Educ Period Obs Variable N Mean Std Dev Median Min Max Associate After 1930 HrWage 239 20.03 8.9 18 6 58 WkWage 381 978.2 529.14 900 99 2885 USLHRS 1688 38.38 14.81 40 0 96 Before 1923 HrWage 242 21.42 10.97 19 2.34 82 WkWage 383 973.74 537.2 900 100 2885 USLHRS 1680 38.08 14.02 40 0 99 Bachelors After 2194 HrWage 148 22.78 11.52 20 3.65 68 WkWage 422 1312.42 685.68 1153.84 38.46 2885 USLHRS 1877 39.05 13.85 40 0 99 Before 2354 HrWage 180 22.46 12.52 20 1 72.12 WkWage 501 1283.06 707.99 1145.83 1 2884.61 USLHRS 2069 39.76 13.63 40 0 99 Graduate After 1194 HrWage 56 30.78 18.44 27 8.35 99 WkWage 253 1622.75 751.89 1509.61 0.01 2885 USLHRS 1099 41.75 13.52 40 0 80 Before 1136 HrWage 45 30.64 18.69 25.15 7 90 WkWage 233 1617.55 719.07 1538.46 9 2885 USLHRS 1017 41.81 13.63 40 0 99 No Degree After 7979 HrWage 978 18.11 8.59 16.5 2.17 90 WkWage 1451 912.81 525.9 800 0.01 2885 USLHRS 6796 36.89 16.87 40 0 99 Before 8404 HrWage 1069 18.05 8.68 16.51 2.1 99.99 WkWage 1557 849.73 481.87 750 20 2885 USLHRS 7208 35.51 17.24 40 0 99 92 Table 3.5: Model Fit Criteria Hourly Wage Model Distribution Assumption N = 53,289 Normal Log Normal Gamma Negative Binomial Deviance 3,721,363 3,701,695 9,745 51,831 Scaled Deviance 53,289 53,289 54,862 51,831 Pearson Chi-Square 3,721,363 3,701,695 11,468 61,889 Scaled Pearson ?2 53,289 53,289 64,562 61,889 Log Likelihood -188,749 -188,608 -174,657 1,681,416 Full Log Likelihood -188,749 -188,608 -174,657 -175,875 AIC (smaller is better) 377,595 377,302 349,399 351,835 AICC (smaller is better) 377,595 377,303 349,399 351,835 BIC (smaller is better) 378,021 377,684 349,781 352,217 Weekly Wage Model Distribution Assumption N = 95,755 Normal Log Normal Gamma Negative Binomial Deviance 25,756,542,240 25,469,200,992 33,236 101,173 Scaled Deviance 95,755 95,755 100,955 101,173 Pearson Chi-Square 25,756,542,240 25,469,200,992 31,881 98,103 Scaled Pearson ?2 95,755 95,755 96,838 98,103 Log Likelihood -734,454 -733,917 -715,357 512,619,943 Full Log Likelihood -734,454 -733,917 -715,357 -715,022 AIC (smaller is better) 1,468,997 1,467,923 1,430,802 1,430,132 AICC (smaller is better) 1,468,997 1,467,923 1,430,802 1,430,132 BIC (smaller is better) 1,469,414 1,468,339 1,431,219 1,430,549 Employment Category N = 457,926 Intercept Only Baseline Logit Cumulative Logit Cumulative Probit Ordered Logit AIC 1,086,164 1,015,220 1,039,500 1,039,269 1,023,461 SC 1,086,197 1,016,494 1,039,946 1,039,716 1,023,975 -2 Log L 1,086,158 1,014,992 1,039,420 1,039,189 1,023,369 93 Table 3.6: Wage and Employment Differences Summary Panel A: Hourly Income Summary No Degree Associates Bachelors Graduate Before After Before After Before After Before After NonDisabled $9.19 $9.19 $11.21 $11.21 $12.99 $12.99 $16.22 $17.18 DisVeteran $10.39 $10.39 $9.31 $9.31 $12.67 $9.07 $18.32 $19.41 Disabled $8.20 $8.20 $10.00 $10.00 $11.59 $11.59 $17.36 $14.21 Veteran $10.02 $10.02 $12.22 $11.21 $12.55 $12.55 $17.68 $18.73 Panel B: Weekly Income Summary No Degree Associates Bachelors Graduate Before After Before After Before After Before After NonDisabled $392 $408 $491 $510 $668 $694 $822 $854 DisVeteran $392 $408 $358 $372 $668 $694 $822 $854 Disabled $316 $328 $396 $346 $538 $559 $662 $662 Veteran $428 $445 $536 $531 $655 $680 $816 $847 Panel C: Binary Odds of Employment No Degree Associates Bachelors Graduate Before After Before After Before After Before After NonDisabled 1.06 1.06 1.92 1.92 1.92 1.92 2.61 2.74 DisVeteran 0.13 0.13 0.24 0.24 0.24 0.24 0.32 0.34 Disabled 0.14 0.13 0.26 0.24 0.41 0.39 0.64 0.67 Veteran 1.06 1.06 1.92 1.66 1.92 1.45 2.61 2.74 Panel D: Odds UnderEmployed Compared to Unemployed Summary No Degree Associates Bachelors Graduate Before After Before After Before After Before After NonDisabled 0.16 0.16 0.26 0.26 0.20 0.20 0.25 0.25 DisVeteran 0.03 0.03 0.10 0.10 0.03 0.03 0.04 0.04 Disabled 0.04 0.04 0.07 0.07 0.05 0.05 0.10 0.10 Veteran 0.12 0.15 0.19 0.23 0.15 0.18 0.19 0.22 Panel E: Odds Employed Compared to Unemployed Summary No Degree Associates Bachelors Graduate Before After Before After Before After Before After NonDisabled 0.91 0.91 1.66 1.66 1.58 1.64 1.98 2.09 DisVeteran 0.11 0.11 0.20 0.20 0.19 0.20 0.24 0.26 Disabled 0.11 0.11 0.21 0.19 0.34 0.32 0.45 0.44 Veteran 1.00 1.00 1.56 1.56 1.72 1.36 1.68 1.77 Panel F: Odds OverEmployed Compared to Unemployed Summary No Degree Associates Bachelors Graduate Before After Before After Before After Before After NonDisabled 0.02 0.02 0.03 0.04 0.05 0.05 0.09 0.11 DisVeteran 0.00 0.00 0.00 0.01 0.01 0.01 0.01 0.02 Disabled 0.00 0.00 0.00 0.00 0.01 0.00 0.01 0.01 Veteran 0.02 0.02 0.03 0.04 0.04 0.02 0.06 0.07 94 Table 3.7: Trade Industry Wage Distributions Fit Hourly Wage Model Distribution Assumption N = 7708 Normal Log Normal Gamma Negative Binomial Deviance 331,568 331,963 1,132 7,189 Scaled Deviance 7,708 7,708 7,892 7,189 Pearson Chi-Square 331,568 331,963 1,493 9,457 Scaled Pearson ?2 7,708 7,708 10,410 9,457 Log Likelihood -25,434 -25,439 -23,190 175,409 Full Log Likelihood -25,434 -25,439 -23,190 -23,531 AIC (smaller is better) 50,943 50,952 46,466 47,136 AICC (smaller is better) 50,943 50,952 46,467 47,136 BIC (smaller is better) 51,200 51,209 46,765 47,393 Weekly Wage Model Distribution Assumption N = 12,373 Normal Log Normal Gamma Negative Binomial Deviance 3,093,923,175 3,052,102,695 4,822 13,126 Scaled Deviance 12,373 12,373 13,121 13,126 Pearson Chi-Square 3,093,923,175 3,052,102,695 4,845 13,254 Scaled Pearson ?2 12,373 12,373 13,183 13,254 Log Likelihood -94,451 -94,367 -91,009 55,077,843 Full Log Likelihood -94,451 -94,367 -91,009 -90,996 AIC (smaller is better) 188,976 188,808 182,104 182,078 AICC (smaller is better) 188,977 188,808 182,105 182,078 BIC (smaller is better) 189,251 189,083 182,424 182,397 Employment Category N = 54,078 Intercept Only Baseline Logit Cumulative Logit Cumulative Probit Ordered Logit AIC 109,125 104,647 108,135 108,114 105,929 SC 109,152 105,768 108,527 108,506 106,320 -2 Log L 109,119 104,395 108,047 108,026 105,841 95 Table 3.8: Trade Industry Wage and Employment Differences Summary Panel A: Hourly Income Summary No Degree Associates Bachelors Graduate Before After Before After Before After Before After NonDisabled $9.92 $9.92 $10.51 $10.51 $11.84 $11.84 $15.32 $15.32 DisVeteran $9.92 $9.92 $10.51 $28.13 $11.84 $11.84 $15.32 $15.32 Disabled $8.65 $8.65 $8.65 $8.65 $10.32 $6.44 $6.96 $6.96 Veteran $9.92 $9.92 $10.51 $10.51 $11.84 $11.84 $15.32 $15.32 Panel B: Weekly Income Summary No Degree Associates Bachelors Graduate Before After Before After Before After Before After NonDisabled $363.29 $363.29 $414.76 $414.76 $619.93 $619.93 $778.99 $778.99 DisVeteran $363.29 $363.29 $414.76 $414.76 $619.93 $619.93 $778.99 $778.99 Disabled $255.83 $255.83 $292.07 $292.07 $604.20 $604.20 $158.44 $158.44 Veteran $363.29 $363.29 $414.76 $414.76 $619.93 $619.93 $778.99 $778.99 Panel C: Binary Odds of Employment No Degree Associates Bachelors Graduate Before After Before After Before After Before After NonDisabled 2.46 2.46 3.19 2.69 3.46 3.46 3.35 3.35 DisVeteran 0.43 1.62 0.55 1.77 0.60 0.41 0.58 2.20 Disabled 0.96 0.96 1.25 1.05 1.35 1.35 1.31 1.31 Veteran 2.46 2.46 3.19 2.69 3.46 1.27 3.35 3.35 Panel D: Odds UnderEmployed Compared to Unemployed Summary No Degree Associates Bachelors Graduate Before After Before After Before After Before After NonDisabled 0.70 0.70 0.87 0.87 0.70 0.70 0.70 0.70 DisVeteran 0.70 0.70 0.87 0.87 0.70 0.70 0.70 0.70 Disabled 0.70 0.70 0.87 0.87 0.70 0.70 0.70 0.70 Veteran 0.47 0.47 0.59 0.59 0.47 0.47 0.47 0.47 Panel E: Odds Employed Compared to Unemployed Summary No Degree Associates Bachelors Graduate Before After Before After Before After Before After NonDisabled 1.97 1.97 2.47 2.47 2.53 2.53 2.44 2.44 DisVeteran 0.32 1.61 0.40 2.02 0.41 2.07 0.39 2.00 Disabled 0.69 0.69 0.87 0.87 0.89 0.89 0.89 0.89 Veteran 1.97 1.97 2.47 2.47 2.53 0.90 2.44 20.44 Panel F: Odds OverEmployed Compared to Unemployed Summary No Degree Associates Bachelors Graduate Before After Before After Before After Before After NonDisabled 0.03 0.03 0.05 0.05 0.09 0.09 0.08 0.08 DisVeteran 0.01 0.01 0.01 0.01 0.02 0.02 0.01 0.01 Disabled 0.01 0.01 0.02 0.10 0.02 0.02 0.02 0.02 Veteran 0.03 0.03 0.04 0.04 0.06 0.06 0.06 0.06 96 Bibliography Industry Statistics Sampler. (2011, May 3rd). Retrieved Januay 14th, 2013, from http://www.census.gov/econ/industry/hist/h336322.htm Automotive News Data Center. (2012, April 3rd). Retrieved April 7th, 2012, from http://www.autonews.com/ United States International Trade Commission. (2012, October 14). Retrieved October 14, 2012, from http://usitc.gov/ Still Waiting for Recovery. (2013, January 5th). The Economist Magazine. Acemoglu, D., & Angrist, J. D. (2001). Consequences of Employment Protection? The Case of the Americans with Disabilities Act. Journal of Political Economy, 915-957. Alam, M., Taufique, K. R., & Uddin, G. S. (2009). Import Inflows of Bangladesh: The Gravity Model Approach. International Journal of Economics and Finance, 131-140. Anderson, J. E. (2011). The Gravity Model. In Annual Review of Economics Book Series, Volume 3 (pp. 133-160). Asche, F., Gordon, D. V., & Hannesson, R. (2004). Tests For Market Integration and the Law of One Price: The Market For Whitefish in France. Marine Resource Economics, 195-210. Autor, D. H. (2011). The Unsustainable Rise of the Disability Rolls in the United States: Causes, Consequences, and Policy Options. National Bureau of Economic Research Working Paper Series, Paper 17697. Autor, D. H., & Duggan, M. G. (2007). Distinguishing Income from Substitution Effects in Disability Insurance. American Economic Review, 119-124. Baier, S. L., & Bergstrand, J. H. (2006). Do Free Trade Agreements Actually Increase Members' International Trade? Journal of International Economics, 72-95. Barten, A. P. (1993). Consumer Allocation Models: Choice of Functional Form. Empirical Economics, 129-158. Burger, M. J., van Oort, F. G., & Linders, G.-J. M. (2009). On the Specification of the Gravity Model of Trade: Zeros, Excess Zeros and Zero-Inflated Estimation. Spatial Economic Analysis, 167-190. Cameron, A. C., & Trivedi, P. K. (2009). Microeconometrics Methods and Applications. New York: Cambridge University Press. Cheong, I., & Wang, Y.-j. (1999). Korea-U.S. FTA Policy: Prospectus and Analysis. Seoul: Korea Institute for International Economic Policy. Chi, T., & Kilduff, P. P. (2010). An Empirical Investigation of the Determinants of Shifting Patterns of U.S. Apparel Imports Using a Gravity Model Framework. Journal of Fashion Marketing and Management, 501-520. Deaton, A., & Muellbauer, J. (1980). An Almost Ideal Demand System. American Economic Review, 312-326. DeLeire, T. (2000). The Wage and Employment Effects of the Americans with Disabilities Act. The Journal of Human Resources, 639-715. Disdier, A.-C., & Marette, S. (2010). The Combination of Gravity and Welfare Approaches for Evaluating Nontariff Measures. American Journal of Agricultural Economics, 713-726. Evans, J. L. (2012). Follow Up on the Haiti Earned Import Allowance Program. Washington, DC: United States Government Accountability Office. Gates, W. H., & Clinton, W. J. (2010, March 10). Senate Foreign Relations Committee with Former President Clinton and Bill Gates. (U. S. Members, Interviewer) 97 Gausch, J. L., & Kogan, J. (2001). Inventories in Developing Countries: Levels and Determinants, a Red Flag on Competitiveness and Growth. Washington, DC: World Bank. Gunderson, J. M., & Hotchkiss, J. L. (2007). Job Separation Behavior of WOTC Workers: Results from a Unique Case Study. Social Service Review, 317-342. Hamersma, S. (2003). The Work Opportunity and Welfare-to-Work Tax Credits: Participation Rates Among Eligible Workers. National Tax Journal, 725-738. Hamersma, S. (2010). Why Don't Eligible Firms Claim Hiring Subsidies? The Role of Job Duration. Economic Inquiry, 916-934. Heaton, P. (2012). The Effects of Hiring Tax Credits on Employment of Disabled Veterans. Rand Occasional Paper Series, OP-366-OSD. Jolls, C., & Prescott, J. J. (2004). Disaggregating Employment Protection: The Case of Disability Protection. Cambridge, MA: National Bureau of Economic Research. Justice, U. D. (2009, March 25). Americans With Disabilities Act of 1990, As Amended. Retrieved May 21, 2012, from http://www.ada.gov/pubs/adastatute08mark.htm#12112a Keller, W. J., & van Driel, J. (1985). Differential Consumer Demand Systems. European Economic Review, 375-390. Klier, T. H., & Rubenstein, J. M. (2009). Imports of Intermediate Parts in the Auto Industry - A Case Study. Washington, DC: Federal Reserve. Labor, U. D. (2012, April 20). Work Opportunity Tax Credit. Retrieved May 23, 2012, from doleta.gov: http://www.doleta.gov/business/incentives/opptax/ Lamont, O. A., & Thaler, R. H. (2003). Anomalies: The Law of One Price in Financial MarketsAuthor. The Journal of Economic Perspectives, 191-202. Lee, J., & Lee, H. (2005). Feasibility and Economic Effects of a Korea-U.S. FTA. Seoul: Korea Institute for International Economic Policy. Lim?o, N., & Venables, A. J. (2001). Infrastructure, Geographical Disadvantage, Transport Costs, and Trade. The World Bank Economic Review, 451-479. Malik, B. A. (1989). A Country Study: Haiti. Washington, DC: Library of Congress Federal Research Division. Matsuda, T. (2005). Differential Demand Systems: A Further Look at Barten's Synthesis. Sourthern Economic Journal, 607-619. Matthews, C. (2013, January 16th). Time.Com. Retrieved February 16th, 2013, from http://business.time.com/2013/01/16/is-walmarts-buy-americanhire-veterans-initiative- anything-more-than-a-pr-stunt/ Merrick, A. (2013, January 22nd). Yahoo Finance. Retrieved February 10th, 2013, from http://finance.yahoo.com/news/wal-mart-speak-hiring-veterans-063248526.html Neves, P. D. (1994). A Class of Differenetial Demand Systems. Economics Letters, 83-86. O'Conner, M. R. (2013, January 11th). Subsidizing Starvation: How American Tax Dollars are Keeping Arkansas Rice Growers Fat on the Farm and Starvin Millions of Haitians. Foreign Policy. Silva, J. S., & Tenreyro, S. (2006). The Log of Gravity. Review of Economics and Statistics, 88(4), 641-658. Theil, H. (1965). The Information Approach to Demand Analysis. Econometrica, 67-87. U.S. - Korea Free Trade Agreement. (n.d.). Retrieved January 15th, 2013, from Office of the United States Trade Representative: http://www.ustr.gov/trade-agreements/free-trade- agreements/korus-fta 98 U.S. Census Bureau. (2012, June 7th). Current Population Survey. Retrieved June 7th, 2012, from http://dataferrett.census.gov/ Vandenbussche, H., Konings, J., & Springael, L. (1999). Import Diversion Under European Antidumping Policy (Working Paper 7340). Retrievable from National Bureau of Economic Research website: http://www.nber.org/papers/w7340.pdf: National Bureau of Economic Research. Wagner, S. A., & Silveira-Camargos, V. (2011). Decision Model for the Application of Just-in- Sequence. International Journal of Production Research, 5713-5736. Walmart. (2013, January 13th). Walmart Corporate. Retrieved February 16th, 2013, from http://news.walmart.com/executive-viewpoints/a-job-to-do-retails-role-in-an-american- renewal Wire Harness Manufacturer's Association. (2013, February 12th). The Wire Harness Manufacturer's Association. Retrieved February 12th, 2013, from http://whma.org/ Woodyard, C. (2012, September 5th). Hyundai's Labor Union Earns Higher Pay in South Korea. USAToday.com. Wooldridge, J. M. (2002). Econometric Analysis of Cross Section and Panel Data. Boston: MIT Press. Wozny, J., & Daly, K. (2001). GAO-01-329 Work Opportunity Tax Credit: Employers Do Not Appear to Dismiss Employees to Increase Tax Credits. Washington, DC: United States General Accounting Office. 99 Appendix A: Table A1.1: Automotive Wire Harness Related HTS Codes HTS Code Description Tariff 8512.20 Electrical lighting or signaling equipment (excluding articles of heading 8539), windshield wipers, defrosters and demisters, of a kind used for cycles or motor vehicles; parts thereof: Other lighting or visual signaling equipment: 8512.20.2040 Lighting equipment: For the vehicles of subheading 8701.20 or heading 8702, 8703, 8704, 8705 or 8711 0% 8512.20.4040 Visual signaling equipment: For the vehicles of subheading 8701.20 or heading 8702, 8703, 8704, 8705 or 8711 2.5% 8544.30.0000 Ignition wiring sets and other wiring sets of a kind used in vehicles, aircraft or ships 5%^ 8701.20 Road tractors for semi-trailers 8702 Motor vehicles for the transport of ten or more persons, including the driver 8703 Motor cars and other motor vehicles principally designed for the transport of persons (other than those of heading 8702), including station wagons and racing cars 8704 Motor vehicles for the transport of goods 8705 Special purpose motor vehicles, other than those principally designed for the transport of persons or goods (for example, wreckers, mobile cranes, fire fighting vehicles, concrete mixers, road sweepers, spraying vehicles, mobile workshops, mobile radiological units) 8711 Motorcycles (including mopeds) and cycles fitted with an auxiliary motor, with or without side-cars; side-cars ^ Increased to 5.3% in 2013 Figure A1.1: Wire Harness Example (Driver Side Door) 100 Figure A1.2: Korean Wire Harness Imports 0 20 40 60 80 100 120 140 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 M illi ons Korean Wire Harness Imports 101 Equilibrium Change: A1) DK = D(PK) A2) SK = S(P) A3) PK = P ? T, T = 1 + ?, ? = Percent Tariff A4) DK = SK A1?) A1??) DK* = ?PK* A2?) A2??) SK* = ?P* A3?) A3??) PK* = P* + T* A4??) DK* = SK* A5) ?(P* + T*) = ?P* A6) ?(PK* - T*) = ?PK* Importer and Exporter Surplus Changes: P0: $4.10 / 100 Q0: 119 Million PA: $3.74 / 100 Q1: 121 Million P: $3.36 / 100 P0 ? PA = $0.36 / 100 PA ? P = $0.38 / 100 Q1 ? Q0 = 2 Million ?IS = ($0.0036*119M) + ($0.0036*2M*0.5) ?IS = $432,000 ?ES = ($0.0038*119M) + ($0.0038*2M*0.5) ?ES = $456,000 ?Welfare = $888,000 102 Appendix B: Table B2.1: HTS Aggregation Level Example 6102 Women's or girls' overcoats, carcoats, capes, cloaks, anoraks (including ski-jackets), windbreakers and similar articles, knitted or crocheted, other than those of heading 6104: 6102.30 Of man-made fibers: 6102.30.05 00 Containing 25 percent or more by weight of leather (635). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... . . . . 5.3% Other: 6102.30.10 00 Containing 23 percent or more by weight of wool or fine animal hair (435). . . . . . . . . . . . . . . . . . . . . 64.4?/kg +18.8% 6102.30.20 Other.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28.2% 10 Women's (635). . . . . . . . . . . . . . . . . . . . . . . . 20 Girls' (635). . . . . . . . . . . . . . . . . . . . . . . . . . . Table B2.2: Endogeneity Tests Percent Tariff Unit Tariff Commodity F Stat Prob > F F Stat Prob > F General 52.23 0.0000 8.45 0.0037 Cotton 1.50 0.2200 - - Animal 5.53 0.0187 3.58 0.0586 103 Table B2.3: Cotton Apparel by Commodity Type Parameter Coat SE Sweater SE T Shirt SE Intercept 12.49 3.52 ** -3.87 3.47 21.44 4.04 ** lgdp 0.39 0.10 ** 0.50 0.14 ** 0.76 0.10 ** lgdpop -1.37 0.17 ** -0.22 0.18 -1.47 0.17 ** LPI 0.03 0.31 -0.47 0.34 -0.38 0.38 ldist 0.11 0.36 0.32 0.33 -1.30 0.43 ** distexcon 0.16 0.02 ** 0.06 0.02 ** 0.18 0.03 ** Perc -0.18 0.03 ** -0.34 0.08 ** -0.30 0.03 ** perc2 0.01 0.00 ** 0.09 0.01 ** 0.01 0.00 ** tradcost 0.01 0.07 -0.12 0.06 -0.40 0.07 ** ExDays 0.00 0.02 0.02 0.02 0.03 0.02 Trade 0.21 0.04 ** 0.23 0.04 ** 0.04 0.06 Y08 0.33 0.28 -0.21 0.27 0.21 0.36 Y09 -0.09 0.29 -0.46 0.28 -0.59 0.37 Y10 -0.50 0.29 -0.12 0.28 -0.63 0.37 Y11 -0.29 0.29 -0.13 0.28 -0.52 0.38 Select Intercept -7.19 2.06 ** -4.75 2.02 * 1.23 . . Select lgdp 0.56 0.06 ** 0.41 0.05 ** 0.40 . . Select lgdpop -0.54 0.08 ** -0.40 0.07 ** -0.71 0.40 Select LPI -0.58 0.17 ** 0.16 0.16 0.22 0.43 Select ldist 0.13 0.25 -0.19 0.22 -0.05 0.52 Select distexcon 0.07 0.02 ** -0.01 0.01 0.00 0.10 Select Perc -0.09 0.02 ** 0.01 0.04 -0.09 0.03 ** Select perc2 0.01 0.00 ** 0.01 0.01 0.00 0.00 Select tradcost -0.14 0.03 ** -0.05 0.03 0.02 0.10 Select lang -0.53 0.15 ** -0.03 0.15 0.00 0.91 Select landlocked 0.18 0.16 0.13 0.17 0.83 1.24 Select ExDocs -0.18 0.04 ** -0.07 0.04 -0.40 0.33 Select Trade 0.08 0.05 0.06 0.07 -0.16 0.06 ** N (OLS / Probit) 116 / 857 114 / 856 AIC 3486 3413 Sigma 2.47 0.11 ** 2.09 0.09 ** Rho -0.75 0.10 ** -0.22 0.30 104 Appendix C: Table C3.1: Hourly Wages (Primary Control Variables) Parameter Estimate SE Pr ?2 Intercept 2.22 0.01 <.0001 GROUP DisVeter 0.12 0.06 0.0324 GROUP Disabled -0.11 0.02 <.0001 GROUP Veteran 0.09 0.01 <.0001 Educ Associate 0.20 0.01 <.0001 Educ Bachelors 0.35 0.01 <.0001 Educ Graduate 0.57 0.01 <.0001 GROUP*Educ DisVeter Associate -0.31 0.13 0.0179 GROUP*Educ DisVeter Bachelors 0.04 0.17 0.8122 GROUP*Educ DisVeter Graduate 0.25 0.22 0.2511 GROUP*Educ Disabled Associate 0.01 0.04 0.8272 GROUP*Educ Disabled Bachelors 0.07 0.05 0.1712 GROUP*Educ Disabled Graduate 0.18 0.09 0.0328 GROUP*Educ Veteran Associate -0.01 0.03 0.7583 GROUP*Educ Veteran Bachelors -0.12 0.03 0.0005 GROUP*Educ Veteran Graduate -0.06 0.07 0.3675 WOTC 0.02 0.02 0.1119 WOTC*GROUP DisVeter -0.14 0.08 0.085 WOTC*GROUP Disabled 0.00 0.03 0.9841 WOTC*GROUP Veteran 0.00 0.02 0.9778 WOTC*Educ Associate 0.00 0.01 0.6757 WOTC*Educ Bachelors 0.01 0.01 0.3108 WOTC*Educ Graduate 0.06 0.02 0.0026 WOTC*GROUP*Educ DisVeter Associate 0.28 0.18 0.1255 WOTC*GROUP*Educ DisVeter Bachelors -0.48 0.23 0.035 WOTC*GROUP*Educ DisVeter Graduate 0.00 0.00 . WOTC*GROUP*Educ Disabled Associate -0.04 0.06 0.5508 WOTC*GROUP*Educ Disabled Bachelors -0.04 0.07 0.5282 WOTC*GROUP*Educ Disabled Graduate -0.26 0.12 0.0345 WOTC*GROUP*Educ Veteran Associate -0.09 0.04 0.0503 WOTC*GROUP*Educ Veteran Bachelors -0.02 0.05 0.731 WOTC*GROUP*Educ Veteran Graduate -0.05 0.09 0.5546 105 Table C3.1 Hourly Wages (Secondary Control Variables) Parameter Estimate SE Pr ?2 WOTC*Sector Industrial -0.03 0.01 0.0127 WOTC*Sector Informatio 0.00 0.03 0.8842 WOTC*Sector Profession -0.01 0.01 0.2498 WOTC*Sector PublicAdm 0.00 0.02 0.8909 WOTC*Sector Services -0.02 0.01 0.1025 AGE 0.01 0.00 <.0001 WOTC*AGE 0.00 0.00 0.7045 RACE 0.06 0.01 <.0001 WOTC*RACE -0.01 0.01 0.1289 SEX 0.16 0.01 <.0001 WOTC*SEX 0.02 0.01 0.0307 Region Midwe 0.02 0.01 0.0026 Region North 0.09 0.01 <.0001 Region West 0.09 0.01 <.0001 WOTC*Region Midwe 0.00 0.01 0.9337 WOTC*Region North -0.01 0.01 0.3515 WOTC*Region West 0.02 0.01 0.0555 106 Table C3.2: Weekly Wages (Primary Control Variables) Parameter Estimate SE Pr ?2 Intercept 5.97 0.01 <.0001 GROUP DisVeter 0.03 0.07 0.7009 GROUP Disabled -0.22 0.02 <.0001 GROUP Veteran 0.09 0.02 <.0001 Educ Associate 0.22 0.01 <.0001 Educ Bachelors 0.53 0.01 <.0001 Educ Graduate 0.74 0.01 <.0001 GROUP*Educ DisVeter Associate -0.32 0.15 0.031 GROUP*Educ DisVeter Bachelors 0.06 0.16 0.7013 GROUP*Educ DisVeter Graduate 0.01 0.21 0.9448 GROUP*Educ Disabled Associate -0.09 0.05 0.0758 GROUP*Educ Disabled Bachelors 0.06 0.05 0.1657 GROUP*Educ Disabled Graduate 0.06 0.06 0.3094 GROUP*Educ Veteran Associate -0.07 0.03 0.0407 GROUP*Educ Veteran Bachelors -0.11 0.03 0.0003 GROUP*Educ Veteran Graduate -0.09 0.04 0.0209 WOTC 0.04 0.02 0.0197 WOTC*GROUP DisVeter -0.13 0.10 0.1748 WOTC*GROUP Disabled -0.01 0.03 0.7829 WOTC*GROUP Veteran 0.03 0.02 0.1251 WOTC*Educ Associate -0.02 0.01 0.0617 WOTC*Educ Bachelors -0.02 0.01 0.0699 WOTC*Educ Graduate -0.01 0.01 0.6049 WOTC*GROUP*Educ DisVeter Associate 0.38 0.21 0.0704 WOTC*GROUP*Educ DisVeter Bachelors -0.31 0.22 0.1658 WOTC*GROUP*Educ DisVeter Graduate -0.13 0.29 0.6572 WOTC*GROUP*Educ Disabled Associate 0.11 0.07 0.1491 WOTC*GROUP*Educ Disabled Bachelors 0.01 0.07 0.8853 WOTC*GROUP*Educ Disabled Graduate -0.09 0.09 0.3050 WOTC*GROUP*Educ Veteran Associate -0.05 0.05 0.3119 WOTC*GROUP*Educ Veteran Bachelors -0.03 0.04 0.4776 WOTC*GROUP*Educ Veteran Graduate -0.05 0.06 0.3928 107 Table C32: Weekly Wages (Secondary Control Variables) Parameter Estimate SE Pr ?2 Sector Industrial 0.16 0.01 <.0001 Sector Informatio 0.22 0.02 <.0001 Sector Profession 0.09 0.01 <.0001 Sector PublicAdm 0.21 0.01 <.0001 Sector Services -0.16 0.01 <.0001 WOTC*Sector Industrial -0.01 0.01 0.5932 WOTC*Sector Informatio -0.01 0.03 0.5843 WOTC*Sector Profession -0.01 0.01 0.3659 WOTC*Sector PublicAdm -0.02 0.02 0.3398 WOTC*Sector Services -0.03 0.02 0.0954 AGE 0.01 0.00 <.0001 WOTC*AGE 0.00 0.00 0.4711 RACE 0.08 0.01 <.0001 WOTC*RACE -0.01 0.01 0.5203 SEX 0.31 0.01 <.0001 WOTC*SEX 0.00 0.01 0.5488 Region Midwe -0.04 0.01 <.0001 Region North 0.05 0.01 <.0001 Region West 0.03 0.01 <.0001 WOTC*Region Midwe 0.00 0.01 0.7536 WOTC*Region North -0.01 0.01 0.4052 WOTC*Region West 0.02 0.01 0.1158 108 Table C3.3: Binary Unemployment (Primary Control Variables) Parameter Estimate SE Pr ?2 Intercept 0.05 0.01 0.0001 GROUP DisVeter -2.10 0.07 <.0001 GROUP Disabled -2.00 0.02 <.0001 GROUP Veteran 0.05 0.03 0.0629 Educ Associate 0.60 0.02 <.0001 Educ Bachelors 0.60 0.01 <.0001 Educ Graduate 0.90 0.02 <.0001 GROUP*Educ DisVeter Associate 0.00 0.16 0.9919 GROUP*Educ DisVeter Bachelors 0.06 0.18 0.7404 GROUP*Educ DisVeter Graduate 0.37 0.26 0.1611 GROUP*Educ Disabled Associate 0.06 0.06 0.3386 GROUP*Educ Disabled Bachelors 0.46 0.05 <.0001 GROUP*Educ Disabled Graduate 0.59 0.08 <.0001 GROUP*Educ Veteran Associate -0.15 0.06 0.0183 GROUP*Educ Veteran Bachelors 0.00 0.06 0.9973 GROUP*Educ Veteran Graduate -0.21 0.09 0.0152 WOTC 0.04 0.02 0.0733 WOTC*GROUP DisVeter 0.01 0.10 0.8865 WOTC*GROUP Disabled -0.06 0.03 0.0448 WOTC*GROUP Veteran 0.02 0.04 0.6722 WOTC*Educ Associate -0.03 0.02 0.2629 WOTC*Educ Bachelors 0.03 0.02 0.1104 WOTC*Educ Graduate 0.05 0.02 0.0396 WOTC*GROUP*Educ DisVeter Associate -0.14 0.22 0.5233 WOTC*GROUP*Educ DisVeter Bachelors 0.36 0.26 0.1679 WOTC*GROUP*Educ DisVeter Graduate 0.16 0.37 0.6684 WOTC*GROUP*Educ Disabled Associate -0.04 0.09 0.6273 WOTC*GROUP*Educ Disabled Bachelors -0.11 0.08 0.1701 WOTC*GROUP*Educ Disabled Graduate -0.08 0.12 0.5104 WOTC*GROUP*Educ Veteran Associate -0.06 0.09 0.5361 WOTC*GROUP*Educ Veteran Bachelors -0.28 0.08 0.0009 WOTC*GROUP*Educ Veteran Graduate 0.14 0.12 0.2644 109 Table C3.3: Binary Unemployment (Secondary Control Variables) Parameter Estimate SE Pr ?2 AGE 0.01 0.00 <.0001 WOTC*AGE 0.00 0.00 0.0476 RACE 0.22 0.01 <.0001 WOTC*RACE -0.03 0.02 0.053 SEX 0.59 0.01 <.0001 WOTC*SEX 0.06 0.01 <.0001 Region Midwe 0.11 0.01 <.0001 Region North 0.00 0.01 0.9165 Region West -0.10 0.01 <.0001 WOTC*Region Midwe 0.01 0.02 0.4305 WOTC*Region North 0.01 0.02 0.5265 WOTC*Region West 0.00 0.02 0.9765 110 Table C3.4: Employment Category (Primary Control Variables - Before) Parameter Status Estimate SE Pr > ?2 Intercept Employed -0.09 0.01 <.0001 Intercept OverEmploy -4.09 0.04 <.0001 Intercept UnderEmplo -1.81 0.03 <.0001 GROUP DisVeter Employed -2.11 0.08 <.0001 GROUP DisVeter OverEmploy -1.93 0.19 <.0001 GROUP DisVeter UnderEmplo -1.87 0.23 <.0001 GROUP Disabled Employed -2.08 0.03 <.0001 GROUP Disabled OverEmploy -2.04 0.08 <.0001 GROUP Disabled UnderEmplo -1.37 0.05 <.0001 GROUP Veteran Employed 0.08 0.03 0.0014 GROUP Veteran OverEmploy 0.08 0.05 0.1235 GROUP Veteran UnderEmplo -0.29 0.07 <.0001 Educ Associate Employed 0.60 0.02 <.0001 Educ Associate OverEmploy 0.63 0.04 <.0001 Educ Associate UnderEmplo 0.47 0.03 <.0001 Educ Bachelors Employed 0.55 0.01 <.0001 Educ Bachelors OverEmploy 1.02 0.02 <.0001 Educ Bachelors UnderEmplo 0.22 0.02 <.0001 Educ Graduate Employed 0.77 0.02 <.0001 Educ Graduate OverEmploy 1.73 0.03 <.0001 Educ Graduate UnderEmplo 0.42 0.03 <.0001 GROUP*Educ DisVeter Associate Employed -0.17 0.19 0.3812 GROUP*Educ DisVeter Associate OverEmploy 0.07 0.42 0.869 GROUP*Educ DisVeter Associate UnderEmplo 0.94 0.37 0.0115 GROUP*Educ DisVeter Bachelors Employed 0.01 0.20 0.9431 GROUP*Educ DisVeter Bachelors OverEmploy -0.33 0.47 0.4816 GROUP*Educ DisVeter Bachelors UnderEmplo 0.52 0.51 0.3116 GROUP*Educ DisVeter Graduate Employed 0.47 0.29 0.1015 GROUP*Educ DisVeter Graduate OverEmploy -0.30 0.57 0.6046 GROUP*Educ DisVeter Graduate UnderEmplo -5.89 17.69 0.739 GROUP*Educ Disabled Associate Employed 0.08 0.07 0.2231 GROUP*Educ Disabled Associate OverEmploy 0.01 0.20 0.9456 GROUP*Educ Disabled Associate UnderEmplo 0.04 0.12 0.7547 GROUP*Educ Disabled Bachelors Employed 0.53 0.06 <.0001 GROUP*Educ Disabled Bachelors OverEmploy 0.24 0.16 0.1353 GROUP*Educ Disabled Bachelors UnderEmplo 0.30 0.12 0.0143 GROUP*Educ Disabled Graduate Employed 0.59 0.09 <.0001 GROUP*Educ Disabled Graduate OverEmploy 0.26 0.20 0.1958 GROUP*Educ Disabled Graduate UnderEmplo 0.51 0.18 0.0041 GROUP*Educ Veteran Associate Employed -0.15 0.07 0.0213 GROUP*Educ Veteran Associate OverEmploy -0.21 0.12 0.0885 GROUP*Educ Veteran Associate UnderEmplo -0.22 0.15 0.144 GROUP*Educ Veteran Bachelors Employed 0.01 0.06 0.8791 GROUP*Educ Veteran Bachelors OverEmploy -0.26 0.11 0.013 GROUP*Educ Veteran Bachelors UnderEmplo -0.02 0.15 0.8741 GROUP*Educ Veteran Graduate Employed -0.25 0.09 0.0047 GROUP*Educ Veteran Graduate OverEmploy -0.47 0.13 0.0002 GROUP*Educ Veteran Graduate UnderEmplo -0.24 0.21 0.2694 111 Table C3.4: Employment Category (Primary Control Variables - After) Parameter Status Estimate SE Pr ?2 WOTC Employed 0.03 0.02 0.2073 WOTC OverEmploy 0.11 0.05 0.0388 WOTC UnderEmplo 0.07 0.04 0.0834 WOTC*GROUP DisVeter Employed 0.05 0.11 0.6238 WOTC*GROUP DisVeter OverEmploy -0.27 0.29 0.3467 WOTC*GROUP DisVeter UnderEmplo -0.32 0.35 0.3573 WOTC*GROUP Disabled Employed -0.08 0.04 0.0329 WOTC*GROUP Disabled OverEmploy -0.30 0.12 0.0103 WOTC*GROUP Disabled UnderEmplo 0.03 0.07 0.6491 WOTC*GROUP Veteran Employed -0.02 0.04 0.6537 WOTC*GROUP Veteran OverEmploy 0.08 0.07 0.2365 WOTC*GROUP Veteran UnderEmplo 0.18 0.09 0.0437 WOTC*Educ Associate Employed -0.02 0.02 0.3994 WOTC*Educ Associate OverEmploy -0.03 0.05 0.5964 WOTC*Educ Associate UnderEmplo -0.03 0.04 0.4825 WOTC*Educ Bachelors Employed 0.04 0.02 0.039 WOTC*Educ Bachelors OverEmploy -0.10 0.03 0.0033 WOTC*Educ Bachelors UnderEmplo 0.06 0.03 0.0624 WOTC*Educ Graduate Employed 0.05 0.02 0.0277 WOTC*Educ Graduate OverEmploy -0.02 0.04 0.5537 WOTC*Educ Graduate UnderEmplo 0.07 0.04 0.1348 WOTC*GROUP*Educ DisVeter Associate Employed 0.06 0.25 0.8177 WOTC*GROUP*Educ DisVeter Associate OverEmploy -0.60 0.65 0.3541 WOTC*GROUP*Educ DisVeter Associate UnderEmplo -0.41 0.58 0.4815 WOTC*GROUP*Educ DisVeter Bachelors Employed 0.41 0.28 0.1465 WOTC*GROUP*Educ DisVeter Bachelors OverEmploy -0.16 0.78 0.8359 WOTC*GROUP*Educ DisVeter Bachelors UnderEmplo 1.08 0.68 0.1112 WOTC*GROUP*Educ DisVeter Graduate Employed 0.05 0.40 0.9009 WOTC*GROUP*Educ DisVeter Graduate OverEmploy 0.44 0.81 0.5895 WOTC*GROUP*Educ DisVeter Graduate UnderEmplo 7.05 17.70 0.6906 WOTC*GROUP*Educ Disabled Associate Employed -0.08 0.10 0.4168 WOTC*GROUP*Educ Disabled Associate OverEmploy 0.40 0.28 0.1451 WOTC*GROUP*Educ Disabled Associate UnderEmplo -0.11 0.18 0.5198 WOTC*GROUP*Educ Disabled Bachelors Employed -0.15 0.09 0.0972 WOTC*GROUP*Educ Disabled Bachelors OverEmploy 0.43 0.23 0.0627 WOTC*GROUP*Educ Disabled Bachelors UnderEmplo -0.34 0.19 0.0658 WOTC*GROUP*Educ Disabled Graduate Employed -0.01 0.13 0.9429 WOTC*GROUP*Educ Disabled Graduate OverEmploy 0.16 0.28 0.5624 WOTC*GROUP*Educ Disabled Graduate UnderEmplo 0.00 0.24 0.9909 WOTC*GROUP*Educ Veteran Associate Employed -0.03 0.09 0.7579 WOTC*GROUP*Educ Veteran Associate OverEmploy -0.03 0.17 0.8758 WOTC*GROUP*Educ Veteran Associate UnderEmplo -0.42 0.22 0.0631 WOTC*GROUP*Educ Veteran Bachelors Employed -0.23 0.09 0.0069 WOTC*GROUP*Educ Veteran Bachelors OverEmploy -0.42 0.15 0.0056 WOTC*GROUP*Educ Veteran Bachelors UnderEmplo -0.33 0.21 0.1152 WOTC*GROUP*Educ Veteran Graduate Employed 0.14 0.13 0.2603 WOTC*GROUP*Educ Veteran Graduate OverEmploy 0.08 0.18 0.6409 WOTC*GROUP*Educ Veteran Graduate UnderEmplo -0.02 0.30 0.9492 112 Table C3.4: Employment Category (Secondary Control Variables) Parameter Status Estimate SE Pr > ?2 AGE Employed 0.01 0.00 <.0001 AGE OverEmploy 0.02 0.00 <.0001 AGE UnderEmplo 0.00 0.00 <.0001 WOTC*AGE Employed 0.00 0.00 0.0355 WOTC*AGE OverEmploy 0.00 0.00 0.2908 WOTC*AGE UnderEmplo 0.00 0.00 0.0054 RACE Employed 0.15 0.01 <.0001 RACE OverEmploy 0.65 0.03 <.0001 RACE UnderEmplo 0.27 0.02 <.0001 WOTC*RACE Employed -0.02 0.02 0.3046 WOTC*RACE OverEmploy -0.10 0.04 0.0087 WOTC*RACE UnderEmplo -0.07 0.03 0.0319 SEX Employed 0.57 0.01 <.0001 SEX OverEmploy 1.63 0.02 <.0001 SEX UnderEmplo -0.44 0.02 <.0001 WOTC*SEX Employed 0.06 0.01 <.0001 WOTC*SEX OverEmploy 0.12 0.03 <.0001 WOTC*SEX UnderEmplo 0.04 0.03 0.1281 Region Midwe Employed 0.08 0.01 <.0001 Region Midwe OverEmploy 0.13 0.03 <.0001 Region Midwe UnderEmplo 0.31 0.02 <.0001 Region North Employed -0.02 0.01 0.0564 Region North OverEmploy -0.08 0.03 0.0043 Region North UnderEmplo 0.26 0.02 <.0001 Region West Employed -0.13 0.01 <.0001 Region West OverEmploy -0.16 0.03 <.0001 Region West UnderEmplo 0.22 0.02 <.0001 WOTC*Region Midwe Employed 0.01 0.02 0.4001 WOTC*Region Midwe OverEmploy -0.02 0.04 0.5942 WOTC*Region Midwe UnderEmplo 0.03 0.03 0.3196 WOTC*Region North Employed 0.02 0.02 0.3351 WOTC*Region North OverEmploy -0.03 0.04 0.4070 WOTC*Region North UnderEmplo 0.01 0.03 0.6959 WOTC*Region West Employed -0.01 0.02 0.6460 WOTC*Region West OverEmploy -0.02 0.04 0.6470 WOTC*Region West UnderEmplo 0.04 0.03 0.2446 113 Table C3.5: Trade Industry Hourly Wages Parameter Estimate SE Pr ?2 Intercept 2.30 0.02 <.0001 GROUP DisVeter 0.14 0.14 0.3392 GROUP Disabled -0.14 0.03 <.0001 GROUP Veteran 0.05 0.03 0.1283 Educ Associate 0.06 0.02 0.0046 Educ Bachelors 0.18 0.02 <.0001 Educ Graduate 0.43 0.04 <.0001 GROUP*Educ DisVeter Associate -0.52 0.30 0.089 GROUP*Educ DisVeter Bachelors -0.54 0.41 0.1867 GROUP*Educ DisVeter Graduate 0.00 0.00 . GROUP*Educ Disabled Associate -0.06 0.09 0.5133 GROUP*Educ Disabled Bachelors 0.29 0.15 0.051 GROUP*Educ Disabled Graduate -0.65 0.23 0.0038 GROUP*Educ Veteran Associate -0.04 0.09 0.6517 GROUP*Educ Veteran Bachelors -0.05 0.10 0.6487 GROUP*Educ Veteran Graduate -0.56 0.38 0.1454 WOTC 0.04 0.03 0.1589 WOTC*GROUP DisVeter -0.34 0.26 0.1966 WOTC*GROUP Disabled -0.02 0.05 0.7083 WOTC*GROUP Veteran -0.07 0.05 0.1181 WOTC*Educ Associate 0.01 0.03 0.6487 WOTC*Educ Bachelors 0.09 0.03 0.0022 WOTC*Educ Graduate 0.02 0.06 0.7257 WOTC*GROUP*Educ DisVeter Associate 0.98 0.46 0.0329 WOTC*GROUP*Educ DisVeter Bachelors 0.10 0.53 0.8516 WOTC*GROUP*Educ DisVeter Graduate 0.00 0.00 . WOTC*GROUP*Educ Disabled Associate 0.11 0.16 0.4696 WOTC*GROUP*Educ Disabled Bachelors -0.47 0.19 0.0139 WOTC*GROUP*Educ Disabled Graduate 0.08 0.29 0.7737 WOTC*GROUP*Educ Veteran Associate 0.11 0.13 0.3952 WOTC*GROUP*Educ Veteran Bachelors -0.22 0.15 0.1473 WOTC*GROUP*Educ Veteran Graduate -0.03 0.47 0.9443 AGE 0.01 0.00 <.0001 WOTC*AGE 0.00 0.00 0.2334 RACE 0.06 0.02 0.0002 WOTC*RACE -0.04 0.02 0.0625 SEX 0.16 0.01 <.0001 WOTC*SEX 0.03 0.02 0.0495 Region Midwe 0.02 0.02 0.2078 Region North 0.10 0.02 <.0001 Region West 0.09 0.02 <.0001 WOTC*Region Midwe -0.03 0.02 0.2565 WOTC*Region North -0.06 0.03 0.0276 WOTC*Region West -0.02 0.02 0.4898 114 Table C3.6: Trade Industry Weekly Wages Parameter Estimate SE Pr ?2 Intercept 5.90 0.03 <.0001 GROUP DisVeter -0.26 0.21 0.2199 GROUP Disabled -0.35 0.05 <.0001 GROUP Veteran -0.02 0.04 0.6148 Educ Associate 0.13 0.03 <.0001 Educ Bachelors 0.53 0.02 <.0001 Educ Graduate 0.76 0.04 <.0001 GROUP*Educ DisVeter Associate -0.61 0.48 0.2045 GROUP*Educ DisVeter Bachelors -0.62 0.48 0.1982 GROUP*Educ DisVeter Graduate 0.00 0.00 . GROUP*Educ Disabled Associate 0.07 0.14 0.6269 GROUP*Educ Disabled Bachelors 0.33 0.16 0.0484 GROUP*Educ Disabled Graduate -1.24 0.36 0.0005 GROUP*Educ Veteran Associate 0.06 0.11 0.6077 GROUP*Educ Veteran Bachelors -0.20 0.12 0.0965 GROUP*Educ Veteran Graduate 0.21 0.22 0.3361 WOTC 0.02 0.04 0.6741 WOTC*GROUP DisVeter 0.03 0.33 0.9354 WOTC*GROUP Disabled 0.01 0.07 0.8392 WOTC*GROUP Veteran 0.08 0.06 0.1817 WOTC*Educ Associate 0.01 0.04 0.8834 WOTC*Educ Bachelors 0.02 0.03 0.4501 WOTC*Educ Graduate -0.03 0.05 0.6263 WOTC*GROUP*Educ DisVeter Associate 0.95 0.64 0.1382 WOTC*GROUP*Educ DisVeter Bachelors -1.31 0.69 0.059 WOTC*GROUP*Educ DisVeter Graduate 0.00 0.00 . WOTC*GROUP*Educ Disabled Associate 0.31 0.20 0.1229 WOTC*GROUP*Educ Disabled Bachelors -0.25 0.22 0.2651 WOTC*GROUP*Educ Disabled Graduate 0.56 0.44 0.2044 WOTC*GROUP*Educ Veteran Associate -0.13 0.16 0.4313 WOTC*GROUP*Educ Veteran Bachelors -0.01 0.17 0.9371 WOTC*GROUP*Educ Veteran Graduate -0.31 0.32 0.3316 AGE 0.01 0.00 <.0001 WOTC*AGE 0.00 0.00 0.7727 RACE 0.14 0.02 <.0001 WOTC*RACE 0.00 0.03 0.9254 SEX 0.36 0.02 <.0001 WOTC*SEX 0.03 0.02 0.1989 Region Midwe -0.02 0.02 0.3431 Region North 0.03 0.02 0.1374 Region West 0.03 0.02 0.1742 WOTC*Region Midwe -0.03 0.03 0.2678 WOTC*Region North 0.00 0.03 0.9641 WOTC*Region West 0.01 0.03 0.8522 115 Table C3.7: Binary Unemployment (Primary Control Variables) Parameter Estimate SE Pr ?2 Intercept 0.90 0.05 <.0001 GROUP DisVeter -1.75 0.26 <.0001 GROUP Disabled -0.94 0.08 <.0001 GROUP Veteran -0.15 0.09 0.0842 Educ Associate 0.26 0.06 <.0001 Educ Bachelors 0.34 0.04 <.0001 Educ Graduate 0.31 0.09 0.0004 GROUP*Educ DisVeter Associate 1.61 0.80 0.044 GROUP*Educ DisVeter Bachelors -0.24 0.60 0.688 GROUP*Educ DisVeter Graduate 0.00 0.00 . GROUP*Educ Disabled Associate 0.14 0.24 0.5403 GROUP*Educ Disabled Bachelors -0.46 0.22 0.0347 GROUP*Educ Disabled Graduate 0.39 0.67 0.563 GROUP*Educ Veteran Associate 0.00 0.24 0.9853 GROUP*Educ Veteran Bachelors 0.54 0.32 0.0865 GROUP*Educ Veteran Graduate -0.36 0.44 0.411 WOTC 0.08 0.07 0.2428 WOTC*GROUP DisVeter 1.33 0.44 0.0025 WOTC*GROUP Disabled 0.10 0.11 0.3715 WOTC*GROUP Veteran 0.10 0.13 0.413 WOTC*Educ Associate -0.17 0.08 0.0392 WOTC*Educ Bachelors -0.09 0.06 0.1317 WOTC*Educ Graduate 0.05 0.12 0.6744 WOTC*GROUP*Educ DisVeter Associate -1.74 1.01 0.0858 WOTC*GROUP*Educ DisVeter Bachelors -1.70 0.85 0.0449 WOTC*GROUP*Educ DisVeter Graduate 0.00 0.00 . WOTC*GROUP*Educ Disabled Associate -0.36 0.33 0.2813 WOTC*GROUP*Educ Disabled Bachelors 0.16 0.32 0.6205 WOTC*GROUP*Educ Disabled Graduate 0.26 0.88 0.7658 WOTC*GROUP*Educ Veteran Associate -0.08 0.34 0.8092 WOTC*GROUP*Educ Veteran Bachelors -1.00 0.38 0.0082 WOTC*GROUP*Educ Veteran Graduate 2.10 1.11 0.0588 116 Table C3.7: Binary Unemployment (Secondary Control Variables) Parameter Estimate SE Pr ?2 AGE 0.01 0.00 <.0001 WOTC*AGE 0.00 0.00 0.0872 RACE 0.30 0.04 <.0001 WOTC*RACE -0.12 0.06 0.0356 SEX 0.51 0.03 <.0001 WOTC*SEX -0.02 0.04 0.6056 Region Midwe -0.09 0.04 0.0407 Region North -0.17 0.04 0.0001 Region West -0.07 0.04 0.0809 WOTC*Region Midwe 0.20 0.06 0.0009 WOTC*Region North 0.13 0.06 0.0475 WOTC*Region West 0.08 0.06 0.1702 117 Table C3.8: Trade Industry Employment Category (Primary Control Variables - Before) Parameter Status Estimate SE Pr ?2 Intercept Employed 0.68 0.05 <.0001 Intercept OverEmploy -3.39 0.11 <.0001 Intercept UnderEmplo -0.36 0.08 <.0001 GROUP DisVeter Employed -1.83 0.28 <.0001 GROUP DisVeter OverEmploy -1.69 0.61 0.0055 GROUP DisVeter UnderEmplo -0.88 0.48 0.0699 GROUP Disabled Employed -1.05 0.08 <.0001 GROUP Disabled OverEmploy -1.28 0.20 <.0001 GROUP Disabled UnderEmplo -0.19 0.11 0.0873 GROUP Veteran Employed -0.10 0.09 0.2762 GROUP Veteran OverEmploy -0.30 0.15 0.0506 GROUP Veteran UnderEmplo -0.39 0.18 0.0294 Educ Associate Employed 0.23 0.06 0.0001 Educ Associate OverEmploy 0.48 0.10 <.0001 Educ Associate UnderEmplo 0.22 0.09 0.0118 Educ Bachelors Employed 0.25 0.05 <.0001 Educ Bachelors OverEmploy 0.93 0.07 <.0001 Educ Bachelors UnderEmplo 0.06 0.07 0.3936 Educ Graduate Employed 0.22 0.09 0.0168 Educ Graduate OverEmploy 0.85 0.13 <.0001 Educ Graduate UnderEmplo 0.05 0.14 0.6899 GROUP*Educ DisVeter Associate Employed 1.31 0.85 0.1245 GROUP*Educ DisVeter Associate OverEmploy -11.89 609.40 0.9844 GROUP*Educ DisVeter Associate UnderEmplo 3.03 0.94 0.0013 GROUP*Educ DisVeter Bachelors Employed -1.85 1.12 0.0979 GROUP*Educ DisVeter Bachelors OverEmploy 0.06 0.94 0.9505 GROUP*Educ DisVeter Bachelors UnderEmplo 1.26 0.86 0.1432 GROUP*Educ DisVeter Graduate Employed 0.00 . . GROUP*Educ DisVeter Graduate OverEmploy 0.00 . . GROUP*Educ DisVeter Graduate UnderEmplo 0.00 . . GROUP*Educ Disabled Associate Employed 0.22 0.25 0.3843 GROUP*Educ Disabled Associate OverEmploy -0.72 0.77 0.3459 GROUP*Educ Disabled Associate UnderEmplo 0.18 0.33 0.578 GROUP*Educ Disabled Bachelors Employed -0.28 0.23 0.217 GROUP*Educ Disabled Bachelors OverEmploy -0.34 0.47 0.4726 GROUP*Educ Disabled Bachelors UnderEmplo -1.52 0.54 0.005 GROUP*Educ Disabled Graduate Employed 0.47 0.71 0.5053 GROUP*Educ Disabled Graduate OverEmploy -9.47 149.90 0.9497 GROUP*Educ Disabled Graduate UnderEmplo 0.80 0.84 0.3414 GROUP*Educ Veteran Associate Employed 0.01 0.25 0.9757 GROUP*Educ Veteran Associate OverEmploy 0.04 0.37 0.9197 GROUP*Educ Veteran Associate UnderEmplo -0.92 0.64 0.1535 GROUP*Educ Veteran Bachelors Employed 0.59 0.32 0.0669 GROUP*Educ Veteran Bachelors OverEmploy -0.59 0.51 0.244 GROUP*Educ Veteran Bachelors UnderEmplo 0.75 0.51 0.1424 GROUP*Educ Veteran Graduate Employed -0.57 0.46 0.2131 GROUP*Educ Veteran Graduate OverEmploy -0.04 0.58 0.9471 GROUP*Educ Veteran Graduate UnderEmplo -0.62 1.09 0.571 118 Table C3.8: Trade Industry Employment Category (Primary Control Variables - After) Parameter Status Estimate SE Pr ?2 WOTC Employed 0.14 0.08 0.0614 WOTC OverEmploy -0.05 0.16 0.7746 WOTC UnderEmplo -0.12 0.11 0.2783 WOTC*GROUP DisVeter Employed 1.63 0.46 0.0004 WOTC*GROUP DisVeter OverEmploy -9.08 82.43 0.9123 WOTC*GROUP DisVeter UnderEmplo -0.24 1.16 0.8379 WOTC*GROUP Disabled Employed 0.13 0.12 0.2894 WOTC*GROUP Disabled OverEmploy -0.17 0.32 0.6022 WOTC*GROUP Disabled UnderEmplo 0.09 0.17 0.598 WOTC*GROUP Veteran Employed 0.03 0.13 0.7968 WOTC*GROUP Veteran OverEmploy 0.30 0.21 0.1534 WOTC*GROUP Veteran UnderEmplo 0.40 0.24 0.1003 WOTC*Educ Associate Employed -0.16 0.08 0.0572 WOTC*Educ Associate OverEmploy -0.27 0.14 0.0513 WOTC*Educ Associate UnderEmplo -0.06 0.12 0.6276 WOTC*Educ Bachelors Employed -0.10 0.06 0.129 WOTC*Educ Bachelors OverEmploy -0.18 0.10 0.0616 WOTC*Educ Bachelors UnderEmplo -0.09 0.10 0.3763 WOTC*Educ Graduate Employed -0.02 0.13 0.8562 WOTC*Educ Graduate OverEmploy 0.25 0.18 0.1686 WOTC*Educ Graduate UnderEmplo 0.18 0.18 0.3325 WOTC*GROUP*Educ DisVeter Associate Employed -1.58 1.06 0.1363 WOTC*GROUP*Educ DisVeter Associate OverEmploy 11.74 625.90 0.985 WOTC*GROUP*Educ DisVeter Associate UnderEmplo -1.18 1.57 0.4499 WOTC*GROUP*Educ DisVeter Bachelors Employed -10.84 86.26 0.9 WOTC*GROUP*Educ DisVeter Bachelors OverEmploy -2.43 237.40 0.9918 WOTC*GROUP*Educ DisVeter Bachelors UnderEmplo 0.19 1.45 0.8937 WOTC*GROUP*Educ DisVeter Graduate Employed 0.00 . . WOTC*GROUP*Educ DisVeter Graduate OverEmploy 0.00 . . WOTC*GROUP*Educ DisVeter Graduate UnderEmplo 0.00 . . WOTC*GROUP*Educ Disabled Associate Employed -0.53 0.36 0.1356 WOTC*GROUP*Educ Disabled Associate OverEmploy 1.90 0.89 0.0335 WOTC*GROUP*Educ Disabled Associate UnderEmplo -0.85 0.51 0.097 WOTC*GROUP*Educ Disabled Bachelors Employed 0.14 0.34 0.6659 WOTC*GROUP*Educ Disabled Bachelors OverEmploy 0.64 0.66 0.3316 WOTC*GROUP*Educ Disabled Bachelors UnderEmplo 0.67 0.71 0.3454 WOTC*GROUP*Educ Disabled Graduate Employed 0.11 0.93 0.9045 WOTC*GROUP*Educ Disabled Graduate OverEmploy -0.44 219.80 0.9984 WOTC*GROUP*Educ Disabled Graduate UnderEmplo 0.37 1.05 0.7276 WOTC*GROUP*Educ Veteran Associate Employed -0.06 0.34 0.8609 WOTC*GROUP*Educ Veteran Associate OverEmploy 0.14 0.51 0.7822 WOTC*GROUP*Educ Veteran Associate UnderEmplo 0.75 0.75 0.3204 WOTC*GROUP*Educ Veteran Bachelors Employed -1.03 0.39 0.0078 WOTC*GROUP*Educ Veteran Bachelors OverEmploy -0.34 0.60 0.5793 WOTC*GROUP*Educ Veteran Bachelors UnderEmplo -0.54 0.62 0.3859 WOTC*GROUP*Educ Veteran Graduate Employed 2.13 1.13 0.0603 WOTC*GROUP*Educ Veteran Graduate OverEmploy 1.34 1.26 0.2883 WOTC*GROUP*Educ Veteran Graduate UnderEmplo 3.00 1.55 0.0531 119 Table C3.8: Trade Industry Employment Category (Primary Control Variables) Parameter Status Estimate SE Pr ?2 AGE Employed 0.01 0.00 <.0001 AGE OverEmploy 0.04 0.00 <.0001 AGE UnderEmplo -0.01 0.00 <.0001 WOTC*AGE Employed -0.01 0.00 0.0246 WOTC*AGE OverEmploy 0.00 0.00 0.8352 WOTC*AGE UnderEmplo 0.00 0.00 0.6602 RACE Employed 0.26 0.04 <.0001 RACE OverEmploy 0.75 0.09 <.0001 RACE UnderEmplo 0.09 0.06 0.1648 WOTC*RACE Employed -0.14 0.06 0.0165 WOTC*RACE OverEmploy -0.07 0.12 0.5476 WOTC*RACE UnderEmplo -0.06 0.09 0.5103 SEX Employed 0.49 0.03 <.0001 SEX OverEmploy 1.66 0.06 <.0001 SEX UnderEmplo -0.64 0.05 <.0001 WOTC*SEX Employed -0.03 0.05 0.4855 WOTC*SEX OverEmploy 0.05 0.09 0.5969 WOTC*SEX UnderEmplo -0.08 0.07 0.2962 Region Midwe Employed -0.07 0.04 0.0983 Region Midwe OverEmploy -0.30 0.08 <.0001 Region Midwe UnderEmplo -0.08 0.07 0.2715 Region North Employed -0.19 0.05 <.0001 Region North OverEmploy -0.24 0.08 0.0016 Region North UnderEmplo -0.01 0.07 0.8514 Region West Employed -0.05 0.04 0.2092 Region West OverEmploy -0.28 0.07 0.0002 Region West UnderEmplo 0.04 0.06 0.5173 WOTC*Region Midwe Employed 0.17 0.06 0.0078 WOTC*Region Midwe OverEmploy 0.28 0.11 0.0081 WOTC*Region Midwe UnderEmplo 0.41 0.10 <.0001 WOTC*Region North Employed 0.13 0.07 0.0556 WOTC*Region North OverEmploy 0.00 0.11 0.9734 WOTC*Region North UnderEmplo 0.21 0.10 0.0401 WOTC*Region West Employed 0.05 0.06 0.4472 WOTC*Region West OverEmploy 0.02 0.11 0.8272 WOTC*Region West UnderEmplo 0.27 0.09 0.0035 120 Table C3.9: Employment Category Comparisons Between Overall Economy and Trade Panel A: Overall Economy Overall Economy Before Overall Economy After Status DisVeter Disabled NonDisab Veteran Total DisVeter Disabled NonDisab Veteran Total Employed Frequency 315 3,237 132,953 8,183 144,688 358 2,898 131,601 7,710 142,567 Overall % 0.1% 1.2% 50.1% 3.1% 54.5% 0.1% 1.1% 50.6% 3.0% 54.8% Row % 0.2% 2.2% 91.9% 5.7% 0.3% 2.0% 92.3% 5.4% Column% 21.0% 18.7% 56.8% 63.1% 23.6% 17.4% 57.4% 62.2% OverEmployed Frequency 46 298 12,602 947 13,893 35 278 13,089 1,045 14,447 Overall % 0.0% 0.1% 4.7% 0.4% 5.2% 0.0% 0.1% 5.0% 0.4% 5.6% Row % 0.3% 2.1% 90.7% 6.8% 0.2% 1.9% 90.6% 7.2% Column% 3.1% 1.7% 5.4% 7.3% 2.3% 1.7% 5.7% 8.4% UnderEmployed Frequency 38 731 16,360 438 17,567 34 696 15,861 450 17,041 Overall % 0.0% 0.3% 6.2% 0.2% 6.6% 0.0% 0.3% 6.1% 0.2% 6.6% Row % 0.2% 4.2% 93.1% 2.5% 0.2% 4.1% 93.1% 2.6% Column% 2.5% 4.2% 7.0% 3.4% 2.2% 4.2% 6.9% 3.6% Unemployed Frequency 1,102 13,021 71,994 3,396 89,513 1,092 12,773 68,936 3,194 85,995 Overall % 0.4% 4.9% 27.1% 1.3% 33.7% 0.4% 4.9% 26.5% 1.2% 33.1% Row % 1.2% 14.6% 80.4% 3.8% 1.3% 14.9% 80.2% 3.7% Column% 73.4% 75.3% 30.8% 26.2% 71.9% 76.7% 30.0% 25.8% Total 1,501 17,287 233,909 12,964 265,661 1,519 16,645 229,487 12,399 260,050 0.57 6.51 88.05 4.88 100 0.58 6.4 88.25 4.77 100 121 Table C3.9: Employment Category Comparisons Between Overall Economy and Trade Panel B: Trade Industry Trade Industry Before Trade Industry After Status DisVeter Disabled NonDisab Veteran Total DisVeter Disabled NonDisab Veteran Total Employed 29 476 16,404 866 17,775 49 413 15,830 817 17,109 0.1% 1.7% 59.8% 3.2% 64.8% 0.2% 1.6% 59.4% 3.1% 64.2% 0.2% 2.7% 92.3% 4.9% 0.3% 2.4% 92.5% 4.8% 32.2% 46.3% 65.4% 70.8% 55.7% 47.2% 64.8% 65.2% OverEmployed 6 37 1,784 98 1,925 0 38 1,776 130 1,944 0.0% 0.1% 6.5% 0.4% 7.0% 0.0% 0.1% 6.7% 0.5% 7.3% 0.3% 1.9% 92.7% 5.1% 0.0% 2.0% 91.4% 6.7% 6.7% 3.6% 7.1% 8.0% 0.0% 4.3% 7.3% 10.4% UnderEmployed 15 144 2,363 52 2,574 12 121 2,291 87 2,511 0.1% 0.5% 8.6% 0.2% 9.4% 0.1% 0.5% 8.6% 0.3% 9.4% 0.6% 5.6% 91.8% 2.0% 0.5% 4.8% 91.2% 3.5% 16.7% 14.0% 9.4% 4.3% 13.6% 13.8% 9.4% 6.9% Unemployed 40 371 4,551 208 5,170 27 304 4,519 220 5,070 0.2% 1.4% 16.6% 0.8% 18.8% 0.1% 1.1% 17.0% 0.8% 19.0% 0.8% 7.2% 88.0% 4.0% 0.5% 6.0% 89.1% 4.3% 44.4% 36.1% 18.1% 17.0% 30.7% 34.7% 18.5% 17.5% Total 90 1,028 25,102 1,224 27,444 88 876 24,416 1,254 26,634 % 0.3% 3.8% 91.5% 4.5% 100.0% 0.3% 3.3% 91.7% 4.7% 100.0% 122 Figure C3.1: Hourly Wage Distribution Figure C3.2: Weekly Wage Distribution 123 Figure C3.3: Usual Hours per Week Figure C3.4: Employment