Do state economic development incentives create jobs? An analysis of state employment tax credits.
Abstract:
Employment tax credits have become one of the primary tools of state economic development policy. A recurring question is whether these credits create jobs that would not have been created in their absence. This paper provides estimates of the employment impact of such credits by comparing the employment change in eligible firms that participate in employment tax credit programs with eligible firms that do not participate in such programs.

Results from a switching regression model indicate that firms taking Georgia's Jobs Tax Credit created 23 to 28 percent more jobs than eligible firms not taking the credit between 1993 and 1995. The cost per job is $2280 to $2680 over the 1993 to 1995 period. While the maximum number of jobs potentially attributable to the program is small, the cost per job is also low especially when compared with firm-specific incentive packages.

Subject:
Tax incentives (Laws, regulations and rules)
Tax credits (Laws, regulations and rules)
Employment tax credit (Laws, regulations and rules)
Author:
Faulk, Dagney
Pub Date:
06/01/2002
Publication:
Name: National Tax Journal Publisher: National Tax Association Audience: Academic; Professional Format: Magazine/Journal Subject: Business; Economics; Law Copyright: COPYRIGHT 2002 National Tax Association ISSN: 0028-0283
Issue:
Date: June, 2002 Source Volume: 55 Source Issue: 2
Geographic:
Geographic Scope: United States Geographic Code: 1USA United States; 0JSTA States
Accession Number:
90119604
Full Text:
INTRODUCTION

Employment tax credits have become one of the primary tools of state economic development policy, with just over one-half of the states in the U.S. offering this form of tax incentive. A recurring question is whether these credits create jobs that would not have been created in their absence. (1) Few studies of this issue have focused on specific state tax incentives as opposed to taxes in general. Yet, policy makers must know such impacts before they can design effective incentive programs. This paper provides estimates of such impacts by comparing the employment change in eligible firms (2) that participate in employment tax credit programs with eligible firms that do not participate in such programs.

Numerous studies have shown that taxes, in general, have a small or no effect on employment. Bartik (1991) and Wasylenko (1997) discuss studies that have examined this relationship. The findings that taxes have a very limited effect on employment may reflect limitations of the tax measure. Since these studies address average effects of all programs operating in a geographic area, the response of individual firms to specific tax incentive programs cannot be teased out. To assess the effect of these tax incentives on employment, firm-level data is needed so that conclusions can be drawn about the effects of employment tax credits on individual firms.

The scant evidence (consisting of one study) on state employment tax credits suggests that these credits modestly influence firms' employment decisions. Using data on business establishments applying for Ohio's Job Creation Tax Credit, Gabe and Kraybill (1999) find that the credit had a positive impact on job creation in Ohio between 1993 and 1995, with between 63 and 68 percent of new jobs (2764 to 3976 jobs) occurring in firms that received the credit (among firms that were eligible to take the credit).

Evidence from federal employment tax credits suggests much smaller effects. These credits have had low participation rates and low impacts on employment. Bishop and Montgomery (1993) estimate that less than 5 percent of firms participated in the Targeted Job Tax Credit program and that at least 70 percent of the tax credits were for workers that would have been hired in the absence of the credit. Perloff and Wachter (1979) estimate that firms which knew about the New Job Tax Credit, a less targeted employment tax credit program in place from 1977-78, created 3 percent more jobs than other firms.

The present analysis differs from previous studies on employment tax credits in two ways. First, it uses firm-level data taken from corporate income tax returns rather than survey data, which are subject to respondent bias and which may inflate the estimates of the employment impact. In addition to being more objective than surveys, these data allow an examination of how firm characteristics affect program participation and employment. Previous work on employment tax credit programs, such as Hamermesh (1976, 1978, 1993), has focused on how wage subsidies affect the quantity of labor demanded, (3) where an employment tax credit serves to reduce wage payments. These studies focus on the elasticity of demand for labor and ignore the problem of employers' failure to participate in the subsidy program.

Second, the empirical model jointly estimates the decision to participate in a tax credit program and the effect of such a tax credit on employment. To achieve this end, a switching regression is used. (4) The model incorporates the decision to participate in a tax credit program into the firm's employment decision. Data from Georgia's Job Tax Credit (JTC) are used to estimate the model. The parameter estimates are then used to calculate the difference in the employment change for participating and nonparticipating firms and thereby provide estimates of the maximum possible effect of the credit on employment in participating firms. The evidence presented in this paper suggests that firms are creating jobs in response to state employment tax credits. Firms taking Georgia's JTC between 1993 and 1995 created 23 to 28 percent more jobs than eligible firms not taking the credit.

The remainder of the paper is organized as follows. The next section presents a brief overview of Georgia's JTC program. The following section provides a framework for analyzing the participation decision and the employment impact of state employment tax credits. The fourth section develops an empirical model and presents estimates of the employment impact of state employment tax credits. The final section offers conclusions.

GEORGIA'S JOB TAX CREDIT

The structure of the JTC influences its economic impact and effectiveness in creating jobs. Some relevant economic features of Georgia's Job Tax Credit include the following:

(1) Georgia's JTC is a tax credit available for the creation of new full-time jobs. This provision may discourage over-time or part-time work since the credit is not available for these types of employment.

(2) The JTC is a credit against corporate income tax liability. Firms with no tax liability cannot use the credit. All firms would qualify for a credit against income tax withholdings, payroll, or the social security contribution, for example.

(3) Fifty percent of a firm's income tax liability is the maximum JTC a firm can take in any year.

(4) If the minimum number of jobs is not maintained, the firm does not have to refund a portion of the previous years' credit back to the state i.e., there are no clawback provisions.

(5) The minimum number of new jobs that a business establishment must create to qualify for the JTC and the credit per job differs depending on the Tier designation of the county in which the establishment is located (Table A-1 in the Appendix). (5) Establishments located in Tier 1 counties have to create fewer jobs to qualify for the JTC, and the credit per job is higher relative to the other tiers. Through this mechanism, the JTC targets business establishments in less developed counties.

(6) An establishment must maintain a minimum increase of jobs for two full years before it can take the JTC. A firm's increase in employment is determined by the increase in the average monthly employment over the firm's fiscal year. The firm supplies this information on the tax credit schedule when claiming the credit.

(7) The JTC can be taken for five years if the jobs are maintained. For example, if a firm in an eligible industry in a Tier 1 county chose 1992 as the base year, created 10 new full-time jobs in 1993 and maintained them in 1994, it can claim a JTC of $20,000 on its 1994 tax return as long as the firm has a state corporate income tax liability of at least $40,000. If the establishment maintains these ten jobs, it can continue to take the JTC through the 1998 tax year The firm must maintain the minimum employment increase for seven years in order to take the credit for five years. The above provisions reduce the incentive for churning i.e., where establishments hire workers, fire them, and then hire new workers to continuously take advantage of the credit.

(8) With the exception of firms in Tier 1 counties, the JTC is limited to certain industries, currently manufacturing and distribution, warehousing, goods processing, tourism, research and development, and information processing. In Tier 1 counties, firms in any industry could take the JTC.

(9) The JTC is nonrefundable, but unused JTC can be carried forward for up to ten years.

(10) Firms can file a Notice of Intent to maintain Tier status. For example, a firm would file a Notice of Intent for a participating establishment located in a Tier 1 county so that if the county is classified as Tier 2 the following year, the firm can continue to claim the Tier 1 credit amount. This provision allows the firm to claim the same credit amount per job created regardless of changes in the Tier designation of the county where the establishment is located. From 1993 to 1995, 29 establishments filed a Notice of Intent for the JTC in Georgia.

(11) The JTC does not require firms to sign an a priori contract guaranteeing the creation of a set number of jobs. Firms can track employment and decide to claim the JTC after the legislated number of jobs is created.

(12) Multi-establishment firms in qualifying industries may take the JTC for jobs created in any single establishment that meets the job creation criteria.

THEORETICAL FRAMEWORK

Faulk (1998) estimates that only 19 percent of eligible firms apply for Georgia's JTC. The decision to participate in an employment tax credit is related to the benefits and costs of participation, which can be linked to certain characteristics of the firm and the structure of the credit. (6)

The Benefit of Participation

The benefit of participation in an employment tax credit program is the value of the tax savings. This is determined by the number of eligible jobs, current and expected tax liability (or past tax liability through net operating loss), the tax credit ceiling, and the discount rate associated with future credits and any carryforward.

The number of creditable jobs and the credit rate determines the maximum credit that a firm can potentially take. This maximum is limited by certain constraints. An income tax credit, especially if it is nonrefundable and nontradable, will have little value to firms with no tax liability. More than 75 percent of corporations in Georgia have no state corporate income tax liability in any given year. (7) The credit ceiling limits the credit to half of tax liability, which also limits the value of the credit. As a result of these constraints, the effective (or average) credit per job may be different from the statutory credit per job. (See Table A-1 for the statutory credit rates.) A firm can take the credit for a number of years if the new jobs are maintained (five years for Georgia's JTC) and can carry forward unused credit for up to ten years, the discount rate associated with the present value of the credit is also a determinant of participation.

The Costs of Participation

Firms incur costs when participating in a tax credit program. These costs fall into six categories: (1) search costs, (2) compliance costs, (3) costs associated with providing additional information to the government, (4) stigma costs, (5) hiring costs, and (6) additional federal tax liability.

Search costs associated with filing employment tax credits include finding out about the credit and other tax abatement programs. Either firms search for ways to reduce their tax liability, or the government can develop a method to notify firms of their eligibility. Because search activities are costly, firms may not continue to search until all tax abatement options are known.

Compliance costs can be divided into two components: startup costs and annual costs. Startup costs include the cost of learning about the credit, training staff, and setting up new forms and systems to capture the information necessary to claim the credit. Annual costs are the year-to-year costs associated with claiming the credit. Firms that participate in an employment tax credit program (or their designated tax advisor) must obtain the necessary forms and gather information needed to flu out the forms. There are also internal coordination costs: within the firm, personnel in charge of hiring decisions need to coordinate activities with personnel in charge of reducing tax liability. For multi-establishment firms, the coordination costs both within and among establishments may become quite substantial.

Costs associated with supplying additional information to the Georgia Department of Revenue may prevent an eligible firm from applying for the credit. Fear of audit (or other consequences of revealing additional information to the Department of Revenue) may be a deterrent. Additional personnel in the Georgia Department of Revenue view the corporate income tax returns of firms that take the JTC. This additional scrutiny may increase the expectation or probability of an audit. In a tax evasion model, Rice (1992)shows that publicly traded companies are more likely to over report income and suggests that these companies do so to avoid audits.

Positive or negative stigma associated with taking the JTC may explain in part why some eligible firms do not file for the credit. The public scrutiny of firms participating in tax abatement programs has increased over the past few years. Recent articles in the popular press have severely criticized tax abatements as a form of corporate welfare. For example, TIME magazine ran a four-part series on corporate welfare (Barlett and Steele, 1998a-d). One might argue that this type of stigma is not apparent with the JTC since minimizing corporate tax liability may be viewed as a good business practice.

Hiring costs are another explanation of the lack of participation in employment tax credit programs. New employees must be interviewed and trained, and the appropriate paperwork must be completed. General Accounting Office (1991) reports that employers participating in the Targeted Job Tax Credit Program estimated that it cost between $600 and $1000 to recruit and train a new employee in the late 1980s. Baron and Bishop (1985) find that hiring costs are positively related to firm size (measured as employment) and the number of establishments within the firm. Because of these differential hiring costs, firms in different industries and of different sizes may find it more or less advantageous to participate in employment tax credit programs. In the case of the JTC, such hiring costs might be larger than the potential credit and may be compounded because the same "new workers" do not have to be employed over the life of the credit. If labor turnover is high, the firm has to replace workers in order to maintain eligibility for the credit.

The deductibility of state corporate income tax liability from federal corporate income tax liability dampens the value of the credit and may be viewed as another cost. The magnitude of the increase in federal income tax liability resulting from state income tax credits depends on the firm's federal corporate tax rate.

The Employment Impact of Participation

It is the firm's responsiveness to a reduction in wages that determines the effectiveness of an employment tax credit in changing a firm's demand for labor. The level of tax liability and the credit ceiling affect the employment impact of the credit. The credit is limited to half of state income tax liability. Because of this credit ceiling, the effective credit per job may be a small portion of the statutory credit, and the reduction in labor costs attributable to the credit may be small. Through this constraint, tax liability ultimately dictates the degree to which the credit reduces the price of labor relative to other factors of production. Changing these structural aspects may enhance the effectiveness of the credit.

EMPIRICAL MODEL AND DATA

Although the benefits and costs of participation and the theoretical determinants of the employment impact of tax credits can be identified, these factors are hard to measure. In addition, it is difficult to estimate the effects of economic development programs because it is difficult to determine what would have happened without the program, i.e., there is no counterfactual. Bartik (1991) suggests using micro data on assisted businesses and a control group of unassisted businesses to examine the effects of specific programs. This study uses tax and employment data for eligible firms that participated and did not participate in Georgia's Job Tax Credit program between 1993 and 1995. (8) The model variables can be linked to the benefits and costs of participation.

To implement the model empirically, a switching regression model is used. The switching regression model is a simultaneous system of three equations: two employment equations and a participation equation that serves as the "switch." Maddala (1983) provides an overview of switching regression models. The equations of the switching regression model are:

[1] [y.sub.1i] = [[beta].sub.1] [x.sub.i] + [u.sub.1i] Employment equation for participants

[2] [y.sub.2i] = [[beta].sub.2] [x.sub.i] + [u.sub.2i] Employment equation for eligible nonparticipants

[3] [y.sub.3i]* = [[gamma].sub.3] [z.sub.i] + [u.sub.3i] Participation equation

where

[y.sub.3i] = 1 iff [y.sub.3i]* > 0

[y.sub.3i] = 0 otherwise.

The error structure for the switching regression model:

[4] [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where [OMEGA] = Cov([u.sub.1i], [u.sub.2i], [u.sub.3i]) =

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

In equations [1] and [2], [y.sub.1i] and [y.sub.2i] are the annual change in employment in eligible firms that participate and do not participate in Georgia's JTC program. In equation [3], [y.sub.3i]* is an unobserved latent variable representing a firm's propensity to participate in the JTC program, and [y.sub.3i] is a dichotomous variable that indicates a firm's participation choice. The [x.sub.i] and [z.sub.i] are vectors of explanatory variables. As discussed previously, the propensity to participate in the JTC program is derived from the benefit of participation less any associated costs. The advantages of this specification are that it treats participation as endogenous and allows the effects of the explanatory variables to differ for participating and nonparticipating firms. The employment equations determine whether or not a firm's participation in the JTC program affects the level of employment. The participation equation (equation [3]) indicates a firm's decision to take the JTC, and this information is used to test for and correct sample selection bias in the employment equation.

Sample selection bias occurs because the same unobservable factors that influence a firm's participation in the tax credit program may influence the likelihood of being in the sample for which the estimates are calculated. Under self-selection those firms that have a comparative advantage in taking the JTC will participate in the program and thus would benefit from it more than would a randomly selected firm with the same characteristics. For example, participating firms may be more likely to increase employment and therefore be more likely to be in the sample.

Heckman's two-step method is used to estimate the model. In short, this method consists of estimating the participation equation with a probit model, using the estimates from the probit to calculate an inverse mills ratio (IMR), and including the IMR as a regressor in each of the employment equations. Heteroskedasticity results from the use of Heckman's two-step method of estimation and is corrected using GLS with the appropriate estimated variance.

Model Specification

The estimating equations are shown below. The variable sources and definitions are shown in Tables A-2 and A-3 of the Appendix.

[5] EMPLOYMENT CHANGE = [[beta].sub.0] +[[beta].sub.1] INITIAL EMPLOYMENT + [[beta].sub.2] PLANTS + [[beta].sub.3] PREVIOUS PARTICIPATION + [[beta].sub.4] START UP + [[beta].sub.5] AGE + [[beta].sub.6] RANK + [[beta].sub.7] IMR + [epsilon]

[6] PARTICIPATION = [[gamma].sub.0] + [[gamma].sub.1] TAX LIABILITY + [[gamma].sub.2] TIER 1 DUMMY + [[gamma].sub.3] TIER 2 DUMMY + [[gamma].sub.4] INITIAL EMPLOYMENT + [[gamma].sub.5] PREVIOUS PARTICIPATION + [[gamma].sub.6] EFFECTIVE JTC + [[gamma].sub.7] HEADQUARTERS LOCATION + [[gamma].sub.8] MANUFACTURING DUMMY + [[gamma].sub.9] START UP + [[gamma].sub.10] RANK*YEAR95 + [epsilon]

Equation [5] is estimated separately for participating and nonparticipating firms. (9)

Data

Data from the firms' corporate income tax returns and establishment-level data from the Georgia Department of Labor's ES202 dataset were used to create the model variables. (10) Nonparticipating firms were selected randomly from a list of eligible firms identified using ES202 data. The dataset consists of 151 firms that were eligible to take the JTC. (11) Seventy of the firms participated in the JTC program, and 81 firms were eligible but did not participate in the program. (12)

The average participating firm had an employment increase of 68 workers (Table 1). Employment increase ranged from -35 to 483. The -35 may seem counterintuitive. Consider a firm that increases employment by 50 workers but only needs to create 10 new jobs to qualify for the credit. If this firm reduces employment by 35 workers, it still qualifies to take a credit for 15 workers. Firms with an initial employment of zero are startups. Startups represent 12 percent of the sample of participating firms. For firms participating in the JTC program, 58 percent had previously participated.

The summary statistics (Tables 1 and 2) show that the average change in employment is larger for nonparticipating firms due, in part, to one firm that had a large change in employment. On average initial employment for nonparticipating firms is higher than for participating firms. Only 6 percent of the nonparticipating firms had participated in the JTC program previously while 37 percent of the sample are startups. Just over a quarter of the firms in the sample have no tax liability.

As shown in Table 3, 46 percent of the firms in the sample participated in Georgia's JTC program between 1993 and 1995. Half of the firms in the sample had locations in Tier 1 counties, and just over 56 percent of the firms were headquartered in Georgia. Just over a third of the firms had participated in the JTC program in previous years. The vast majority of the firms (89 percent) were in manufacturing industries. A quarter of the sample was startups in the base year. The average tax liability of firms in the sample is just over $302,000.

Employment Equation

Employment change is measured as the annual increase in employment for the firm. For nonparticipating firms, this variable was calculated from ES202 data. For participating firms, it was taken from the JTC schedule included with the corporate income tax return. A more detailed breakdown of the average and total employment change in participating and nonparticipating firms is shown in Table 4. The data indicates that the average and total employment change in both participating and nonparticipating firms increased steadily from 1993 to 1995. (13)

Initial employment is included as a measure of firm size. The relationship between employment change and firm size has been widely debated in the literature. Wagner (1992) provides an overview. Larger firms are expected to have a higher level of employment change relative to smaller firms.

The number of plants is also a measure of firm size. Wasylenko (1981) suggests that multi-establishment firms are more responsive to tax incentives because they can locate capital (labor) intensive plants in low tax jurisdictions.

The dummy variable for previous participation in the JTC program is included to examine how taking the JTC in previous years affects current employment. Of particular interest is whether firms that took the JTC in the past continue to increase employment.

It might be argued that startup firms should be modeled separately since the startup decision is different from the decision to expand employment in an existing plant. However, given the number of observations in the sample (39 of 151 firms are startups), a separate model is not feasible. The dummy variable for startup firms is included to distinguish between firms that startup in the base year and existing firms that expand.

The age variable is included to determine if younger firms have a greater change in employment than older firms do.

The rank variable measures a county's level of economic development and differences in the value of the JTC. Bartik (1991) suggests that jobs created in areas with high unemployment are more valuable than jobs created in areas with low unemployment and argues that for state and local incentives to produce national benefits, places with higher unemployment should offer greater incentives than places with low unemployment so that jobs are redistributed from places with lower unemployment to places with higher unemployment. If the JTC induces such a change, establishments located in less developed counties--those counties with highest unemployment and poverty rates, the lowest manufacturing wage and per capita income--would be more likely to create jobs in response to the JTC than establishments in other counties. If this is true, the rank variable will be positively related to the change in employment.

The coefficient on the inverse mills ratio (Lambda) estimates the covariance between the error terms of the participation equation and each employment equation and indicates whether or not sample selection bias is evident.

Participation Equation

As discussed earlier, a firm's tax liability determines, in part, the benefit of participation. Firms with a larger tax liability receive a larger effective credit and should be more likely to participate. To provide information on the degree to which firms are able to use the JTC, more detailed information on the available credit, credit actually taken and the effective credit for participating firms is shown in Table 5. The average JTC carryforward from previous years, available credit in the current tax year, and JTC taken were $80,310, $109,870, and $71,548, respectively. On average, firms are able to utilize about 38 percent of the total available JTC in a given tax year.

Firms located in tier 1 counties have a larger statutory credit rate and lower job creation criteria relative to tiers 2 and 3 and therefore may be more likely to take the JTC. See Table A-1 in the Appendix for credit amounts per job and job creation criteria. Five of the firms in the sample contained establishments located in counties that changed tiers during the study period. Four of these counties were reclassified from tier 2 to tier 1, and one was reclassified from tier 1 to tier 2.

Smaller firms may be more likely to participate. First, firms located in less developed counties are more likely to be smaller relative to firms in more developed counties, e.g., the average size of firms in the sample located in tier 1, 2 and 3 counties is 200, 304 and 661 workers, respectively. Second, smaller firms may have greater incentive to participate in the tax credit program. While it is true that smaller firms may have fewer resources to devote to finding tax abatement, it is also true that abated taxes may be a larger proportion of total costs. Also, smaller firms may face credit rationing or other financial constraints that make tax credits more valuable to them. In addition, larger firms may experience greater difficulty coordinating information needed to claim the JTC, which increases the cost of taking the credit. In a smaller firm, the same person is more likely to be in charge of hiring and taxes, so coordination costs are lower. Gabe and Kraybill (1998) and Pope and Kuhle (1996) find that smaller firms are more likely to participate in an employment tax credit program.

Past participation should be a good predictor of current participation. Since firms have already incurred the cost of finding out about the credit and developing the appropriate systems to track information necessary to claim the credit, they should continue to participate.

As discussed earlier, the maximum JTC is limited to half of a firm's tax liability, so the amount of the credit available to the firm is not necessarily directly related to the number of jobs that are creditable in each firm. For example, if a firm in a tier 1 county created 15 jobs and is eligible to take the credit in 1995, the maximum credit that the firm could potentially take is $37,500 (=2500*15). If the firm has a corporate income tax liability of only $10,000, the firm would take a credit for $5000 and carry forward the $32,500 difference. The effective credit (5000/15 = 333) determines the benefit of the job tax credit. A firm with a higher effective credit should be more likely to take the credit since the benefit per job is higher. (14)

Whether a firm is headquartered in the state is used to measure the likelihood that firms have information about tax abatement options and thus face lower participation costs. Detailed information on how to file for the JTC, the credit amounts, the tier structure, job creation criteria, or qualifying industries is not readily available in the Georgia corporate income tax form or instructions. Thus, being in-state may imply better information via word of mouth, better-informed tax advisors, etc.

The startup dummy and manufacturing dummy are included to determine if these types of firms are more likely to participate in the tax credit program. Startups may be more likely participate since they can take the credit for their entire payroll. However, startups typically have low tax liability for the first several years of operation, which may reduce the likelihood of participation.

In 1995 tier 3 firms became eligible to take the JTC. The interaction term is included to control for this change in credit structure.

RESULTS

Determinants of the Change in Employment

The analysis of the employment impact of employment tax credits seeks to determine if both the change in employment and the determinants of employment change are different for participating and nonparticipating firms. Tables 6 and 7 show the parameter estimates for participating and nonparticipating firms, respectively.

Model estimates show a positive relationship between firm size and the change in employment for both participating and nonparticipating firms, indicating that the change in employment is greater in larger firms. However, the magnitude of the effect differs between firms that participate in the JTC and those that do not. Employment growth for firms participating in the JTC program is about 60 percent larger than it is for firms of similar size that do not participate. The magnitude of the parameter estimate is larger for participating firms, suggesting that size has a larger effect on employment in participating firms than in nonparticipating firms. (15)

Being a startup in the base year is not a significant determinant of the change in employment for firms that took the JTC. It is significant for firms that did not take the JTC. Startups that did not take the credit had an average employment change of 162.5 workers, while startups that took the credit had an average employment change of 79.5 workers.

The coefficients on the IMR are negative but insignificant in both samples indicating that sample selection bias does is not present. These coefficients measure the covariance between the error terms in the two employment equations. The insignificance of this coefficient for nonparticipating firms provides some evidence that these firms did not participate in the JTC program because they did not know about the program and did not incur the search costs to find out about the credit rather than their making a conscious decision not to participate. For participating firms, the insignificance of the coefficient provides some evidence that the tax credit was not a driving force behind an average firm's employment decision. Using a similar model, Gabe and Kraybill (1999) find a negative but significant coefficient on the Inverse Mills Ratio for firms receiving Ohio's Job Creation Tax Credit.

In sum, the determinants of the change in employment are different for participating and nonparticipating firms. The differences in the parameter estimates for the two sets of firms suggest that there are structural differences in the growth patterns of participating and nonparticipating firms. The coefficient on the inverse mills ratio is measured with less precision but is smaller for nonparticipating firms relative to participating firms. This suggests that the hiring decision and participation have a lower correlation for non-participating firms relative to participating firms. It may be that nonparticipating firms do not know about the credit. An alternative explanation is that they are unwilling to incur the search costs to find out about the credit; perhaps because they believe that the cost of finding out about tax abatement is substantial relative to the credit. In this case, the firm has a higher tax liability than necessary as a result of not taking the credit. Finally, a nonparticipating firm may know about the credit and not take it because other participation costs are higher than the credit.

Determinants of Participation

The parameter estimates for the participation equation are shown in Table 8, the marginal effects (calculated at the variable mean), which show how the probability of taking the JTC changes when firm characteristics are slightly altered, are shown in Table 9. These results generally support the hypothesis that participation depends on the benefits and costs.

The probit model estimates show that tax liability is a significant, positive but small influence on the firm's likelihood of taking the JTC. As the marginal effects in Table 9 indicate, changes in the pre-credit tax liability of the average firm has a relatively small effect on a firm's probability of taking the JTC. According to these results, a $10,000 increase in tax liability increases the probability of filing for the JTC by 0.5 percentage point.

Even though the credit amount per job is higher and the job creation threshold is lower in less developed counties, there is no evidence that firms located in such counties are more likely to take the credit. The tier level of the county in which the firm is located is not a significant influence on a firm's likelihood of taking the JTC. Even though 68 percent of participating firms are located in Tier 1 (poor) counties, the estimation results show that firms located in less developed counties are not more likely to participate in the JTC program when other factors are taken into account. (16)

Not unexpectedly, firms that previously took the JTC are more likely to take the JTC in the current year. The marginal effects indicate that previous participation has a relatively large effect on a firm's probability of taking the credit. The model estimates indicate that the difference in the parameter estimate for firms previously participating in the JTC and those not participating is 1.68.

Firms headquartered in Georgia are more likely to take the JTC. As the marginal effects indicate, changes in headquarters location has a relatively large effect on a firm's probability of taking the credit.

The manufacturing dummy is not a significant influence on a firm's likelihood of taking the JTC. Even though 89 percent of the sample are manufacturing firms, when other variables are taken into account, this is not a significant determinant of participation. In their study of Ohio's Job Creation Tax Credit, Gabe and Kraybill (1998) use a larger dataset and include 18 industry dummies as explanatory variables to determine if business establishments in certain industries are more likely to receive a tax credit. None of the industry variables are significant. For the Georgia and Ohio employment tax credit programs, at least, industry does not appear to be a significant influence on the likelihood of participation.

Jobs Attributable to the Employment Tax Credit

To evaluate the benefit of the JTC program, we would like to know the number of jobs created as a result of the program. To estimate this, we compare the employment change in firms that participate in the JTC program with similar firms that do not participate in the program. A portion of this difference in the employment change is attributable to the JTC. (17) Maddala (1983) shows two methods of evaluating program impact. Both methods are used here.

The first method is to compare the change in employment, [y.sub.1i], for a participating firm i and the expected employment if the firm had not participated. The last term in both equations is the inverse mills ratio, and the [sigma] terms are the coefficients on the inverse mills ratios. Under the normality assumption, the change in employment due to participation is:

[7] [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

The change in employment potentially attributable to the JTC is the summation over all participants. This calculation subtracts the predicted employment change if participants had not participated from the observed change in employment of participants. According to this calculation, for the 1993-95 period the total number of new jobs attributable to the JTC is 1870. This is 23.5 percent of the employment change in participating firms. (18) For the firms in the sample, the tax expenditure on the JTC over the 1993-95 period was just over $5 million, so the tax expenditure per new job created is $2678 over the 1993-95 period.

Another method is to calculate the expected growth in employment given that each type of firm participates in the tax credit program. This calculation subtracts the predicted employment change for nonparticipants if they had participated in the program from the predicted employment change of participants.

[8] [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

With this calculation, the number of new jobs attributable to the JTC is 2,196, which is 27.6 percent of the employment change in participating firms. The tax expenditure per job is $2,280 over the 1993-95 period.

According to these estimates, between 1,870 and 2,196 jobs or between 23.5 and 27.6 percent of the employment change in participating firms can be attributed to the JTC. The cost per job is between $2,280 and $2,678. If all of the jobs that were credited (7,951 jobs) were actually attributable to the JTC, then the cost per job is $630. Through the JTC, the state reduced corporate income tax liability of participating firms by just over $5 million between 1993 and 1995, but $3.6 million to $3.8 million of this was a credit for jobs that would have been created in the absence of the JTC program.

While the number of jobs attributable to the program is low, the cost per job relative to the cost for other programs is also low. When compared to some of the large incentive packages that states have offered large corporations over the past several years, broadly applied programs such as employment tax credits appear to be as effective in terms of initial job creation as incentive packages offered to entice large corporations to locate in a particular state. For example, the incentive package that the state of Alabama offered Mercedes is estimated to have cost just short of $170,000 per job for 1500 jobs. (19)

CONCLUSIONS

This paper presents an analysis of the effect of a state job tax credit on employment changes in eligible firms. Firm-level data from corporate income tax returns of eligible firms are used in the analysis. Tax liability, past participation, and headquarters location influence participation in the JTC program. Firm size influences the employment change of participating firms. The parameter estimates from a switching regression model are used to estimate the employment impact of the tax credit.

The evidence presented in this study suggests that firms are creating jobs in response to tax incentives such as Georgia's JTC. Firms taking the credit created 23.5 to 27.6 percent more jobs (1870 to 2196 jobs) than eligible firms not taking the credit between 1993 and 1995. The flip side of the issue is that 72.4 to 76.5 percent of the employment change in participating firms would have been created in the absence of the credit. Since the total tax expenditure on the JTC program was just over $5 million during this period, the state gave up $3.6 to $3.8 million on jobs that would have been created without the credit.

APPENDIX

Acknowledgments

I wish to thank Roy Bahl, Chris Bjornson, Shit Gurmu, Florenz Plassmann, Dave Sjoquist, Eric Schansberg, Mark Strazicich, Mary Beth Walker, Douglas Holtz-Eakin, Therese McGuire, and two anonymous referees for their helpful comments and discussions of earlier versions of this paper. Any remaining errors are my own.

(1) A related issue is whether these credits actually lead to the creation of new jobs or the redistribution of jobs from one geographic area or firm to another. This issue is not addressed here.

(2) Eligible firms are those firms in qualifying industries that create the minimum number of new jobs.

(3) Hamermesh (1978) discusses three types of subsidies for jobs: employment subsidies, wage subsidies, and hiring subsidies. Employment subsidies apply for the entire time a worker is with a firm. Wage subsidies are a fixed percentage of wages, a flat dollar amount or a fixed percentage of wages with a maximum. Hiring subsidies offset training and hiring costs for the initial period of employment in a firm. Employment tax credits are a form of wage subsidy.

(4) This approach assumes that firms not participating in the tax credit program would have created more jobs if they had participated.

(5) Each year the Department of Community Affairs ranks the level of economic development in each of Georgia's 159 counties and assigns each county to one of three tiers. A county's level of economic development is determined by its unemployment rate, average manufacturing wage, poverty rate, and per capita income. Tier 1 counties are the least developed, while tier 3 counties are the most developed.

(6) The following analysis assumes that the credit is applied only to corporate tax liability and is nonrefundable, as is the case for the majority of state employment tax credit programs.

(7) Annual Reports, Georgia Department of Revenue.

(8) For multi-establishment firms, data has been aggregated to the firm level.

(9) Model identification deserves further explanation. The tax liability variable and the previous JTC variable essentially identify the model. Tax liability is statistically significant in the participation equation and not included in the employment change equations. The previous JTC variable is significant in the participation equation and insignificant (as a nonlinear transformation of the original variable through the IMR) in the employment equations. As such, nonlinearity aids in the identification of the model. The author would like to thank one of the anonymous referees for providing detailed comments on this issue.

(10) The ES202 dataset contains information on monthly employment levels, industry, unemployment tax payments, and total wage bill and county for each business establishment in Georgia.

(11) Some of the firms are in the sample for more than one year during the 1993-95 period.

(12) This paper addresses the impact of employment tax credits in firms eligible (created the minimum number of jobs necessary to qualify) to take the credit and thereby excludes firms that did not create enough jobs to qualify for the credit. Due to the exclusion of these firms, the estimates of the employment impact may be lower than they would be if noneligible firms had been included in the analysis.

(13) The dramatic increase in the average employment change and total employment change for nonparticipating firms between 1994 and 1995 is the result of one firm that grew by over 2,000 employees.

(14) The author would like to thank one of the anonymous referees for suggesting the use of this variable.

(15) This might seem to conflict with the finding that smaller firms are more likely to take the credit. However, even though larger firms are less likely to take the credit, among firms taking the credit, larger firms create more jobs than larger firms that do not take the credit.

(16) Of the startups, 33 percent (13 firms) are located in Tier 1 counties.

(17) This is an estimate of the maximum (upper bound) effect of the credit on employment in participating firms. In reality, only a portion of the this difference is likely to be attributable to the credit since many other factors influence employment decisions.

(18) The employment change is the number of new jobs reported by a firm taking the JTC. Since some of these jobs would have been created in the absence of the credit, only a portion of these jobs is actually attributable to the JTC.

(19) However, because employment is more concentrated, the multiplier effects associated with a large plant citing are likely to be larger than those associated with a broadly applied employment tax credit.

REFERENCES

Bartik, Timothy J. Who Benefits from State and Local Economic Development Policies? Kalamazoo, MI: W.E. Upjohn Institute, 1991.

Barlett, Donald L., and James B. Steele. "Corporate Welfare." Time (November 9, 1998a): 36-54.

Barlett, Donald L., and James B. Steele. "Fantasy Islands and Other Perfectly Legal Ways That Big Companies Manage to Avoid Billions in Federal Taxes." Time (November 16, 1998b): 79-93.

Barlett, Donald L., and James B. Steele. "Paying a Price for Polluters." Time (November 23, 1998c): 72-82.

Barlett, Donald L., and James B. Steele. "The Empire of the Pigs." Time (November 30, 1998d): 52-66.

Barton, John M., and John Bishop. "Extensive Search, Intensive Search, and Hiring Costs: New Evidence on Employer Hiring Activity." Economic Inquiry 23 No. 3 (July, 1985): 363-82.

Bishop, John H., and Mark Montgomery. "Does the Targeted Job Tax Credit Create Jobs at Subsidized Firms?" Industrial Relations 32 No. 3 (Fall, 1993): 289-306.

Faulk, Dagney. "Georgia's Job Tax Credit: An Analysis of the Characteristics of Eligible Firms." Fiscal Research Program Report No. 8. Atlanta: Georgia State University, June 1998.

Gabe, Todd M., and David S. Kraybill. "Tax Incentives and Offers in a State Economic Development Program." Review of Regional Studies 28 No. 3 (1998): 1-14.

Gabe, Todd M., and David S. Kraybill. "The Effects of Tax Incentives on Enterprise Growth." Department of Agricultural, Environmental and Development Economics, The Ohio State University. Mimeo, 1999.

General Accounting Office. Targeted Jobs Tax Credit Employer Actions to Recruit, Hire, and Retain Eligible Workers Vary. GAO HRD 91-33, Washington, D.C., February 1991.

Hamermesh, Daniel S. Labor Demand. Princeton, NJ: Princeton University Press, 1993.

Hamermesh, Daniel S. "Subsidies for Jobs in the Private Sector." In Creating Jobs: Public Employment Programs and Wage Subsidies, edited by John L. Palmer, 87-122. Washington, D.C.: Brookings Institution, 1978.

Hamermesh, Daniel S. "Econometric Studies of Labor Demand and Their Applications to Policy Analysis." Journal of Human Resources 11 No. 4 (Fall, 1976): 507-25.

Maddala, G. S. Limited Dependent and Qualitative Variables in Econometrics. Cambridge: Cambridge UP, 1983.

Perloff, Jeffrey M., and Michael L. Wachter. "The New Jobs Tax Credit: An Evaluation of the 1977-78 Wage Subsidy Program." American Economic Review 69 No. 2 (May, 1979): 173-9.

Pope, Ralph A., and James L. Kuhle. "Tax Credits for Job Creation and Job Retention in the California Economy." Public Finance Quarterly, 24 No. 2 (April, 1996): 192-215.

Rice, Eric M. "The Corporate Tax Gap: Evidence on Tax Compliance by Small Corporations." In Why People Pay Taxes Tax Compliance and Enforcement, edited by Joel Slemrod, 125-61. Ann Arbor, MI: Michigan University Press, 1992.

Wagner, Joachim. "Firm Size, Firm Growth, and Persistence of Chance: Testing Gibrat's Law with Establishment Data from Lower Saxony, 1978-89." Journal of Small Business Economics 4 No. 2 (June, 1992): 125-31.

Wasylenko, Michael. "Taxation and Economic Development: The State of the Economic Literature." New England Economic Review (March/April, 1997): 37-52.

Wasylenko, Michael. "The Location of Firms: The Role of Taxes and Fiscal Incentives." In Urban Government Finance Emerging Trends, Volume 20, Urban Affairs Annual Review, edited by Roy Bahl, 155-90. Beverly Hills, CA: Sage Publications, 1981.

Employment tax credits have become one of the primary tools of state economic development policy. A recurring question is whether these credits create jobs that would not have been created in their absence. While numerous studies have shown that taxes, in general, have small or insignificant effects on employment, this study focuses on a specific tax incentive, state employment tax credits, and uses firm-level data from state corporate income tax returns.

This paper discusses the benefits and costs of participation in employment tax credit programs and provides estimates of the employment impact of such credits by comparing the employment change in eligible firms that participate in employment tax credit programs with eligible firms that do not participate in such programs. The empirical model jointly estimates the decision to participate in a tax credit program and the effect of such a tax credit on employment using data from firms eligible to take Georgia's Job Tax Credit (JTC).

Evidence presented in this study suggests that firms are creating jobs in response to tax incentives such as Georgia's JTC. Results from a switching regression model indicate that firms taking Georgia's Jobs Tax Credit created 23 to 28 percent more jobs (1870 to 2196 more jobs on average) than eligible firms not taking the credit between 1993 and 1995. The cost per job is $2280 to $2680 over the 1993 to 1995 period. While the maximum number of jobs potentially attributable to the program is small, the cost per job is also low especially when compared with firm--specific incentive packages.

Dagney Faulk Indiana University Southeast, New Albany, IN 47150-6405
TABLE A-1

GEORGIA'S JOB TAX CREDIT ELIGIBILITY REQUIREMENTS, MINIMUM JOB
CREATION CRITERIA, AND CREDIT AMOUNTS PER JOB CREATED

Tax Year     Tier 1 Counties      Tier 2 Counties      Tier 3 Counties

1991         Jobs:   10           Not Eligible         Not Eligible
             Credit: $1,000

1992         Jobs:   10           Not Eligible         Not Eligible
             Credit: $1,000

1993         Jobs:   10           Jobs:   10           Not Eligible
             Credit: $2,000       Credit: $1,000

1994         Jobs:   10           Jobs:   10           Not Eligible
             Credit: $2,000       Credit: $1,000

1995         Jobs:   10           Jobs:   25           Jobs:   50
             Credit: $2,500       Credit: $1,500       Credit: $500

1996         Jobs:   10           Jobs:   25           Jobs:   50
             Credit: $2,500       Credit: $1,500       Credit: $500

1997         Jobs:   10           Jobs:   15           Jobs:   25
             Credit: $2,500       Credit: $1,500       Credit: $500
TABLE A-2

DESCRIPTION OF VARIABLES USED IN THE EMPLOYMENT EQUATION

Variable              Description

Employment            Employment in period t less employment
Change                in period t - 1

Initial Employment    Employment level in the base year within
                      the state of Georgia

Plant                 Number of establishments within firm
                      (tax entity)

Previous JTC          =1 if firm took JTC in a previous year
                      =0 otherwise

Startup               =1 if the base year employment was zero
                      =0 otherwise

Age of firm           Age measured as date of incorporation in
                      Georgia less the income tax year

Rank                  Tier ranking of county where firm is
                      located (This is an indicator of the level of
                      development of the county)

Lambda                Inverse Mills Ratio

Variable              Source

Employment            Georgia Corporate Income Tax
Change                Returns or ES202 data

Initial Employment    Georgia Corporate Income Tax
                      Returns or ES202 data

Plant                 ES202 data

Previous JTC          Georgia Corporate Income Tax
                      Returns

Startup               Georgia Corporate Income Tax
                      Returns or ES202 data

Age of firm           Georgia Corporate Income Tax
                      Returns

Rank                  Georgia Corporate Income Tax
                      Returns or ES202 data

Lambda                Calculated from the
                      participation (probit) equation
TABLE A-3

DESCRIPTION OF VARIABLES USED IN THE PARTICIPATION EQUATION

Variable                        Description

Participation      =1 if firm took the JTC (had a positive JTC)
Dummy              =0 if firm did not take the JTC or claimed
                   zero JTC.

Tax Liability *    Pre JTC tax liability on the Georgia
                   Corporate Income Tax return (in 10,000s).

Tier 1 Dummy       =1 if located in a Tier 1 county.
                   =0 otherwise.

Tier 2 Dummy       =1 if located in a Tier 2 county.
                   =0 otherwise.

Initial            The number of employees in the base year
Employment **

Previous JTC       =1 if firms took JTC previously
                   =0 otherwise

Effective JTC *    The JTC that a firm claims (in participating
                   firms) or can potentially claim (in
                   nonparticipating firms) divided by the
                   number of credited or potentially credited
                   jobs (in 100s).

Headquarters       =1 if firm's headquarters is in Georgia
Location           =0 otherwise

Manufacturing      =1 for manufacturing firms
Dummy              =0 otherwise

Start up           =1 if the base year employment was zero
                   =0 otherwise

Rank x Year 95     Interaction of the rank of the county
                   where firm is located and a year dummy

Variable                    Source

Participation      Georgia Corporate Income
Dummy              Tax Returns and ES202 data

Tax Liability *    Georgia Corporate Income
                   Tax Returns

Tier 1 Dummy       Georgia Corporate Income
                   Tax Returns or ES202 data

Tier 2 Dummy       Georgia Corporate Income
                   Tax Returns or ES202 data

Initial            Georgia Corporate Income
Employment **      Tax Returns or ES202 data

Previous JTC       Georgia Corporate Income
                   Tax Returns

Effective JTC *    Georgia Corporate Income
                   Tax Return and Author's
                   calculation

Headquarters       Georgia Corporate Income
Location           Tax Returns

Manufacturing      Georgia Corporate Income
Dummy              Tax Returns or ES202 data

Start up           Georgia Corporate Income
                   Tax Returns or ES202 data

Rank x Year 95     Author's calculation

* Tax liability and the Effective JTC were scaled by 10,000 and
100, respectively so that their order of magnitude would be
similar to the other variables in the model.

** For multi-establishment firms, annual employment for all
establishments in a particular firm that participates
or is eligible to participate in the JTC program is used.
TABLE 1

SUMMARY STATISTICS OF VARIABLES USED IN THE EMPLOYMENT
EQUATION, PARTICIPATING FIRMS

Variable                  Mean     Standard Deviation

Employment Change       68.2857         100.5274
Initial Employment     250.3714         352.3007
Plant                    5.6857           9.7185
Previous JTC             0.5857           0.4961
Startup                  0.1285           0.3371
Age                     18.6428          17.2954
Rank                    46.4142          39.3461
IMR                      0.4851           0.4881
   Obs. = 70

Variable                Minimum               Maximum

Employment Change           -35                   483
Initial Employment            0                  1259
Plant                         1                    52
Previous JTC                  0                     1
Startup                       0                     1
Age                           0                    67
Rank                          1                   158
IMR                   0.119E-13                  1.70
   Obs. = 70
TABLE 2

SUMMARY STATISTICS OF VARIABLES USED IN THE EMPLOYMENT
EQUATION, NONPARTICIPATING FIRMS

Variable                  Mean     Standard Deviation

Employment Change       99.1604         240.1943
Initial Employment     344.6172         687.7551
Plant                    3.2469           6.7943
Previous JTC             0.0617           0.2421
Startup                  0.3703           0.4859
Age                     19.8888          19.0794
Rank                    74.8168          42.6097
IMR                     -0.4192           0.3992
   Obs. = 70

Variable                Minimum               Maximum

Employment Change           -15                  2062
Initial Employment            0                  4540
Plant                         1                    57
Previous JTC                  0                     1
Startup                       0                     1
Age                           0                    84
Rank                          4                   159
IMR                       -1.95            -0.214E-01
   Obs. = 70
TABLE 3

SUMMARY STATISTICS OF VARIABLES USED IN THE PARTICIPATION EQUATION

                                                          Standard
Variable                                 Mean             Deviation

Participation Dummy                      0.4635             0.5003
Tax Liability ($ ten thousands)         30.2178            82.2216
Tier 1 Dummy                             0.5099             0.5015
Tier 2 Dummy                             0.3509             0.4788
Initial Employment                     300.9271              558.2
Previous JTC                             0.3046             0.4617
Effective JTC (in hundreds)              5.7299             6.9469
Headquarters Location                    0.5695             0.4967
Manufacturing Dummy                      0.8940             0.3088
Start up                                 0.2582             0.4391
Rank * Year 95                          46.2875            51.5242
  Obs.= 151

                                       Minimum             Maximum
                                     (Counts for         (Counts for
Variable                           Dummy Variables)    Dummy Variables)

Participation Dummy                     0(81)               1(70)
Tax Liability ($ ten thousands)             0              517.41
Tier 1 Dummy                            0(74)               1(77)
Tier 2 Dummy                            0(98)               1(53)
Initial Employment                          0                4540
Previous JTC                           0(105)               1(46)
Effective JTC (in hundreds)                 0               39.12
Headquarters Location                   0(65)               1(86)
Manufacturing Dummy                     0(16)              1(135)
Start up                               0(112)               1(39)
Rank * Year 95                              0                 159
   Obs.= 151
TABLE 4

EMPLOYMENT CHANGE BY YEAR AND PARTICIPATION STATUS

                          Average Employment Change

                            1993     1994     1995

Participating Firms         54.8     58.7     82.4
Nonparticipating Firms        27     35.1    123.3

                            Total Employment Change

                            1993     1994     1995

Participating Firms          877     1349     2554
Nonparticipating Firms        54      702     7276
TABLE 5

USE OF GEORGIA'S JTC BY PARTICIPATING FIRMS (IN DOLLARS)

                                     Mean      Std. Dev.       Sum

Carryforward from Previous years     80,310      147,089     5,621,713
Available JTC in Current Year       109,870      143,406     7,690,917
Total JTC Available *               190,180      261,445    13,312,630
JTC Taken in Current Year            71,548      115,804     5,008,399
Effective Credit                        682          639        47,764
Obs.=70
                                        Min        Max

Carryforward from Previous years          0      819,021
Available JTC in Current Year             0      834,332
Total JTC Available *                     0    1,607,858
JTC Taken in Current Year               414      585,125
Effective Credit                          6         3911
Obs.=70

* Carryforward from Previous years plus JTC available in current year.
TABLE 6

PARAMETER ESTIMATES OF THE
EMPLOYMENT MODEL, PARTICIPATING FIRMS

                   Parameter      Standard
Variable           Estimate        Errors     T-ratio

Intercept           69.7694        47.8099     1.459
Initial
   Employment        0.1360 *       0.0365     3.724
Plant               -0.9922         1.3467    -0.737
Previous JTC       -46.3685        36.2462    -1.279
Startup             46.9534        36.1348     1.299
Age                  0.6412         0.7367     0.870
Rank                -0.3116         0.3203    -0.973
IMR                -12.9749        35.0488    -0.370
Obs. = 70        R-sq. = .2627

Note: The final step of Heckman's procedure was
implemented using GLS.

** Significant at the 0.05 level in a two-tailed test.
TABLE 7

PARAMETER ESTIMATES OF THE
EMPLOYMENT MODEL, NONPARTICIPATING
FIRMS

                   Parameter      Standard
Variable           Estimate        Errors     T-ratio

Intercept            6.5957        92.9532     0.071
Initial
   Employment        0.0803 **      0.0444     1.810
Plant                2.5156         4.0337     0.624
Previous JTC       -20.9658       165.7702    -0.126
Startup            116.3356 **     63.1789     1.841
Age                 -1.9772         1.4655    -1.349
Rank                 0.6978         0.7351     0.949
IMR                 -4.8330        113.171    -0.043
Obs. = 81        R-sq. = .1285

Note: The final step of Heckman's procedure was
implemented using GLS.

* Significant at the 0.10 level in a two-tailed test.
TABLE 8

PROBIT PARAMETER ESTIMATES OF THE
PARTICIPATION MODEL

                    Parameter     Standard
Variable            Estimate       Errors      T-ratio

Intercept          -0.0038         0.9693        -0.004
Tax Liability       0.0126 **      0.0056         2.231
Tier 1 Dummy       -0.5197         0.8260        -0.629
Tier 2 Dummy       -0.7742         0.6593        -1.129
Initial
   Employment      -0.0003         0.0004        -0.738
Previous JTC        1.6836 **      0.3421         4.922
Effective JTC       0.0266         0.0246         1.079
Headquarters
   Location         0.7904 **      0.2841         2.781
Manufacturing
   Dummy           -0.4188         0.5118        -0.818
Start up           -0.3862         0.3479        -1.110
Rank/Year 95
   Interaction     -0.0081         0.0052        -1.560
Obs.= 151

Goodness of Fit: The joint predictions for the model
were 71/81 for JTCD = 0 and 53/70 for JTCD = 1. The
total predictions were 88/81 for JTCD = 0 and 63/70
for JTCD = 1.

Log likelihood function = -60.84664

**Significant at the 0.05 level in a two-tailed test.
TABLE 9

MARGINAL EFFECTS OF PARTICIPATION
IN THE JTC PROGRAM

                  Effect on the
                  Probability of    Standard
Variable          Taking the JTC     Errors     T-ratio

Tax Liability         0.0050 **      0.0022        2.243
Tier 1 Dummy         -0.2072         0.3294       -0.629
Tier 2 Dummy         -0.2967         0.2630       -1.128
Initial
   Employment        -0.0001         0.0001       -0.739
Previous JTC          0.6712 **      0.1359        4.939
Effective JTC         0.0106         0.0098        1.078
Headquarters
   Location           0.3151 **      0.1133        2.782
Manufacturing
   Dummy             -0.1669         0.2041       -0.818
Start up             -0.1539         0.1387       -1.109
Rank/Year 95
   Interaction       -0.0032         0.0021       -1.560

Note: Marginal effects are calculated at mean values
of the independent variables.

** Significant at the 0.05 level in a two-tailed test.
Gale Copyright:
Copyright 2002 Gale, Cengage Learning. All rights reserved.