Printer Friendly

An analysis of pull factors and local sales taxation in Georgia counties.

INTRODUCTION

Perhaps partly due to the rapid and substantial appreciation in housing values between 2002 and 2007, a number of state legislatures considered reforms calling for a decrease or elimination of property taxes and the replacement of such revenue with increased retail sales taxes. Based on data from the Federal Housing Finance Agency, examples of over-the-year housing appreciation during this period are twenty-five percent in Florida, twelve percent in Pennsylvania, and twelve percent in some metropolitan statistical areas of Georgia, such as Savannah (Federal Housing Finance Agency, 2008).

In Florida, former Senate President John McKay was leading a proposal that would replace the state-imposed school tax with broader sales tax collections. The amendment was approved by the Florida Taxation and Budget Commission, and was going to be placed on the November 2008 ballot. Voters were to be asked to replace $9.6 billion in property taxes (the portion of local property taxes that fund schools) with a one-cent increase in the state-wide sales tax and new sales taxes. The Florida Tax Reform proposed legislation would have reduced property taxes by 25 percent or more by 2010. However, the proposal was never placed on the ballot because the Florida Supreme Court ordered the issue removed from the November election (Tax Foundation, 2008).

In May 2008, the governor of Iowa signed into law a bill that established a state-wide one cent sales tax for school infrastructure (Iowa Department of Revenue, 2008). This tax replaced the existing special local option penny sales tax for school infrastructure--approved by voters in each of Iowa's 99 counties. Supporters of the tax argue that such a single-penny state tax will distribute the money more equally to school districts across the state, and thus local property taxes could be reduced since such revenue would no longer be required to build and maintain school facilities.

Similarly, Pennsylvania introduced the School Property Tax Elimination Act (SPTEA) in 2007. This legislation would have eliminated all school property taxes across the Commonwealth. However, SPTEA was pending legislation as the session closed, and thus expired with the end of session in late November 2008. The legislation was re-introduced during the 2009 legislative session as HB-1275. The sponsor of the bill argued that the funds would be replaced primarily by a state sales tax (Rohrer, 2009).

More recently, North Dakota's Secretary of State approved a ballot measure to eliminate property taxes in the state effective January 1, 2012 and replace the funds with sales taxes, individual and corporate income taxes, and other sources (Tax Foundation, 2010).

In Georgia, the centerpiece of the 2008 Session of the General Assembly was House Resolution 900. This tax reform proposal included provisions that would scale back or limit the use of property taxes in funding municipal governments (Georgia Chamber of Commerce, 2007). To replace the funding, local governments would then rely on state grants of funds raised through the imposition of taxes on retail sales and services. Although advocates of wholesale restructuring of local government finance have backed away from early versions of the legislation eliminating virtually all property taxes, that goal appears to remain an important motivating factor for legislative tax reform proposals (Atlanta Journal Constitution, 2007).

OBJECTIVE

Given the great attention state lawmakers are allocating to tax reform proposals, in particular, the desire to replace property taxes with retail sales taxes, the purpose of this study is to develop a better understanding of the factors influencing retail sales activity in Georgia counties. A clearer perspective on the determinants of retail sales activity, including retail sales tax differentials across state lines, would enable policy makers to better understand the potential impacts of increasing retail sales taxes in Georgia. To this end, several models are developed that partly explain cross-sectional variation in the county-level pull factors for retail sales. A pull factor is a measure of the strength of retail sales activity in a particular geographic area, usually a city or county. The following hypotheses will be tested:

Hypothesis 1: Sales tax differentials will affect border-county pull factors. In particular, when a county shares a border with a state where the retail sales tax rate is lower, the pull factor for the border county will be lower, all else equal.

Hypothesis 2: County level pull factors will be higher, all else equal, when the county is on the coast, the proportion of elderly in the population increases, and in counties with higher per capita income.

The empirical results highlight the importance of sales tax differentials on shopping location decisions, and the effects of cross-border shopping on local revenue collections, both important considerations for policymakers considering altering the structure of sales taxes.

LITERATURE REVIEW

Retail sales pull factors have been commonly used to describe retail sales activity in small-scale geographic areas. A large volume of the literature on the topic is applied (Kansas Department of Revenue, 2007; Missouri Department of Economic Development, 2008; Richard Caplan & Associates, 2008; Norman Economic Development Coalition, 2009) and tends to focus on descriptive statistics and the interpretation of a city or county level pull factor. However, academics have also rigorously investigated determinants of pull factors (Mikesell, 1970; Yanagida, et. al., 1991; Gale, 1996). The academic strand in the literature tends to focus more on inferential statistics and larger geographic areas (such as a metropolitan statistical area) or a larger volume of geographic units to support the testing of hypotheses.

Turning first to the applied literature, much of the research focuses on the computation of the pull factor for a county (Missouri Department of Economic Development, 2008) or city (Richard Caplan & Associates, 2008; Norman Economic Development Coalition, 2009) as the unit of analysis. The research is generally used to quantify the strength of the area's retail trade and identify weaknesses or leakages of retail trade or other business activity out of the area (Harris, 2003). Upon identifying such a leakage, the analysis may proceed to identify "target" retail or business service sectors that would theoretically thrive in the city or county where the pull factor indicates a "hole" exists (Richard Caplan & Associates, 2008). The identified "target" sectors could relocate or expand capacity in the region and thereby capture retail or business activity that was determined to be leaking out of the area.

A significant drawback of much of the applied literature is the reliance on descriptive statistics and a notable lack of inferential statistics and hypothesis testing. Many times, the research ends with a discussion of what the computed pull factor or related trade area market size happened to be and how they changed through time, and no effort is given to provide a causal explanation for the computed value of the pull factor (Missouri Department of Economic Development, 2009). Other applied research in this vein may anecdotally cite the presence of certain industries and large retail sales establishments to explain why the computed figure happened to be what it is (Richard Caplan & Associates, 2008).

In a manuscript that tends to bridge the gap between applied research and academic research, Hughes (2004) examines the correlation between county level pull factors and causal variables capturing the empirical implications of central place theory (Christaller, 1966) such as population density, interstate access, per capita income, population, elderly population, and commuting patterns.

Indeed, central place theory serves as the primary theoretical foundation for rigorous empirical investigations of the determinants of retail trade pull factors. The hierarchical structure of retail trade markets explains why some goods and services are offered for sale in some locations and not in other locations. Central places will offer a wider range of retail activity than outlying areas (Shaffer et al., 2003).

Making use of central place theory, Yanagida, et.al. (1991) demonstrate that pull factors in agricultural counties in Nebraska are a function of income levels within the community, demographic characteristics, and distances to trade centers. Their findings show that larger pull factors are associated with higher per capita income, greater distance to competing trade centers, and small population declines.

Gale (1996) estimates pull factors for over 3,000 U.S. counties. His findings indicate that pull factors are lower in relatively rural counties than in urbanized counties. Also, county characteristics such as higher population density, county size and interstate highway access were associated with higher pull factors, while regional effects such as West Coast, Northeast, and Midwest were weak, showing no significant impacts on pull factors.

The analytical technique has also been used to examine idiosyncratic features of the regional market for retail trade. For example, Broomhall and King (1995), Stone (1993), Hicks and Wilburn (2001) deployed the technique to examine the impact that a mass merchandiser, Walmart, exerted on retail trade activity in a defined geographic area. Another strand in the pull factor literature explores the effects of tax policy on retail trade activity (Mikesell, 1970; Walsh & Jones, 1988; Stone & Artz, 2002; Nelson, 2002; Tosun & Skidmore, 2007).

Mikesell (1970) examines retail sales in 173 central cities of Standard Metropolitan Statistical Areas and found that a one point increase in the central-city tax relative to suburban tax rate implied a 1.7 percent to 11 percent reduction in central-city per capita sales. The work done by Walsh and Jones (1988) is directed more to the problem of border county retail sales loss. They find that for border counties in West Virginia, each 1 percent reduction in the sales tax rate increased grocery store sales by 5.9 percent.

More recently, Tosun and Skidmore (2007) evaluated the issue of cross-border effects in West Virginia counties over a four-year period from 1988 to 1991. Their results indicate that for every one percentage point increase in county relative price ratio due to the sales tax change, per capita food sales decreased by 1.38 percent. Their research indicates that after a six percent increase in sales tax rate at the beginning of 1990, food sales in West Virginia border counties fell by about eight percent. Lastly, Nelson (2002) shows that in the U.S., determinants of state excise tax policy are strongly influenced by the size of potential cross-border markets. State legislators in states with a large potential cross-border market are more likely to set tax rates below that of neighboring states to attract non-resident customers for specific commodities such as cigarettes, motor fuel, beer and insurance premiums.

This research falls into this latter strand in the literature about pull factors. We use inferential statistics to consider the effects of tax policy on retail trade activity, as measured by pull factors, in which we draw on central place theory as a foundation for the empirical model.

METHOD AND DATA

The pull factor is a measure of a community's retention of sales (Shaffer, 1989) and can be used to characterize the relative strength of a community's retail sector. In this study, the definition of a pull factor is the division of the county's per capital retail sales tax receipts by the state's per capita retail sales tax receipts as follows:

county retail sales tax [receipts.sub.i] / county [population.sub.i] Pull Factor = / state retail sales tax receipts / state population (1)

The ratio estimates the proportion of retail sales in county i that were captured by retail trade establishments in county i, or alternatively, trade activity that did not leak out of county i. A factor greater than one suggests the county is capturing sales from within its own borders while a factor less than one suggests that county sales are leaking to areas outside the county.

Data from 2006 from 159 counties in the state of Georgia is used to estimate a model of county level pull factors (Georgia Department of Revenue, 2006). Ordinary least squares regression analysis with White's (1980) correction for heteroskedasticity was used. Following the literature based on central place theory, explanatory variables include per capita income, percent of the population aged 65 or over, population density, distance to the closest MSA, and other features of the county characterizing its location on the coast or interstate access to border counties of adjacent states.

The model takes the general form of:

[PF.sub.i] = f([PCI.sub.i], [POP65.sub.i], [DENS.sub.i], [HIGH.sub.i], [COAST.sub.i], [TOMSA.sub.i], [HIFL.sub.i], [HIAL.sub.i], [HISC.sub.i], [HITN.sub.i]) (2)

Where:

[PF.sub.i] = pull factor measure for each county, in 2006;

[PCI.sub.i] = per capita income in county i, in 2006;

[POP65.sub.i] = percentage of the population in county i aged 65 years and above;

[DENS.sub.i] = population density of county i, in 2006;

[HIGH.sub.i] = a binary variable equal to 1 if an interstate highway is located in the county and 0 otherwise;

[COAST.sub.i] = a binary variable equal to 1 if county i is a coastal county on the Atlantic Ocean and 0 otherwise;

[TOMSA.sub.i] = miles from the seat of county i to the closest Metropolitan Statistical Area (MSA);

[HIFI.sub.i] = a multiplicative variable combining HIGH and FL. FL is a binary variable equal to 1 if county i borders the state of Florida and 0 otherwise;

[HIAL.sub.i] = a multiplicative variable combining HIGH and AL. AL is a binary variable equal to 1 if county i borders the state of Alabama and 0 otherwise;

[HISC.sub.i] = a multiplicative variable combining HIGH and SC. SC is a binary variable equal to 1 if county i borders the state of South Carolina and 0 otherwise;

[HITN.sub.i] = a multiplicative variable combining HIGH and TN. TN is a binary variable equal to 1 if county i borders the state of Tennessee and 0 otherwise.

The relationships implied by central place theory between the independent variables and the dependent variable are discussed next. Higher levels of per capital income are assumed to support greater retail sales as discretionary income increases. This suggests a positive coefficient on PCI. An increased elderly population base (POP65) is assumed to raise the pull factor because the elderly are less likely, at the margin, to travel extensively to engage in retail trade and therefore these individuals purchase a larger portion of retail goods and services in the county of residence. The expected sign on the coefficient for DENS is positive. In a densely populated region, the agglomerative effects of high density shopping venues imply that the coefficient for DENS is positive. In densely populated regions, agglomerative effects generate the demand and resulting supply of a full range of retail business trade.

A variable characterizing interstate highway access (HIGH) captures the effect of transportation accessibility on retail trade activity. Interstate access could increase or decrease retail trade in a given county because counties characterized by greater access could be more attractive for retail stores, but at the same time such accessibility could make it easier for residents to drive elsewhere to shop. Therefore, the overall effect of highway access on pull factors is uncertain and remains an empirical matter.

The binary variable COAST is included to capture the tourism-based inflow of retail sales activity into coastal counties. As such, COAST is expected to have positive effects on the pull factor of coastal counties. The distance between county i and the closest MSA (TOMSA) is expected to have a positive effect on the pull factor in county i. As the distance to the closest MSA increases, it becomes more costly (requires more travel) for county residents to reach establishments in other retail markets. Thus, counties situated farther away from MSAs are more likely to have higher pull factors and an accordingly lower level of sales leakage.

HIFL and HIAL are included in the model to ascertain whether counties bordering Florida and Alabama with easy cross-border access are likely to experience increased or decreased retail trade. Under the assumption that differential sales tax rates along state borders could possibly affect shopping patterns along Georgia's border, the principal hypothesis is that per capita retail sales will be lower in counties with an adverse tax differential because border-county residents will find it advantageous to shop out of the state and non-residents will find such state shopping less attractive. Since cross-border access is generally good in the major population centers of Georgia bordering the state of Florida, such as Camden, Clinch, Echols and Thomas counties, and such counties have a lower sales tax than the Florida border counties, a positive sign is expected on the independent variable HIFL. A similar analysis for Alabama is conducted with the variable HIAL. In general, Alabama border-counties have a lower sales tax than neighboring Georgia counties; thus, a negative coefficient is expected for this variable.

Although North Carolina borders the state of Georgia, no Georgia county bordering North Carolina has interstate access. South Carolina (SC) and Tennessee (TN) and their multiplicative versions (HISC and HITN) are included to characterize the effects of these state's tax differentials and interstate access on the dependent variable, PF.

Taxable sales data for the retail sector published by the Georgia Department of Revenue for 2006 was used to compute the retail sales pull factors. County level data for PCI and DENS was obtained from the U.S. Census Bureau for the year 2006, and POS65 was collected from the 2000 census (U.S. Census Bureau, 2006). The variable TOMSA was the estimated distance from the county seat to the closest Metropolitan Statistical Area. The variable HIGH was constructed by reviewing the American Map[R], 2006 United States Road Atlas.

Table 1 provides descriptive statistics for each of the variables in the study as they enter the model in their log-transformed state or in their untransformed state. Table 2 shows the correlation matrix for non-binary variables. It suggests the absence of substantial multicolinearity among the primary variables of the model, with the exception of several cases involving PCI discussed in greater detail below.

RESULTS

Following convention in the literature, the model is estimated in a semi-double natural log specification (indicated by "LOG"). The log specification of both the dependent variable and some of the independent variables (semi-double log) allows for the interpretation of the estimated parameters as elasticities indicating the percentage change in the pull factor from a one percent change in the log transformed variables of the right hand side of the equation. Explanatory variables that can take a value of zero, primarily the binary and interactive variables, are not log-transformed because the log of zero is undefined. The variable TOMSA, is not log-transformed because it takes a value of zero for all cases in which the county seat is in a metropolitan statistical area.

Table 3 provides the results from estimating equation (3) by ordinary least squares regression analysis, using the White's (1980) correction for heteroskedasticity. The following equation encompasses all four versions of the estimated models:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (3)

where the variables are as defined above, Cn is the model parameter to be estimated, and [E.sub.i] is a random error term.

Model 1 serves as a baseline model for comparison, from which the binary and interactive variables that do not significantly contribute to the explanatory power of the model are excluded. In Model 2, PCI was omitted to assess the effect on the regression results attributed to the correlation between PCI and DENS. In the third model, HISC and HITN are added to Model 1 to demonstrate the difference in the marginal effects from highway connections to bordering states. Model 4 demonstrates the inadequacy of modeling the effects of border states without considering the transportation access to those states characterized by the interactive state and highway variables.

The t-statistics for the base equation, Model 1, indicate statistical confidence at the 95 percent level or higher for all variables except HIAL. The adjusted R-squared is 0.42, while the F-statistic of 15.82 indicates statistical significance at the one percent level of the model as a whole.

The estimated parameters of Model 1 provide an estimate of the percentage change in the dependent variable if the continuous causal variable changes by one percent, or if binary causal variables change from zero to one. For example, a one percent increase in the population aged 65 and above causes a one-half percent (0.49) increase in the pull factor. Increasing levels of per capita income and density are associated with higher pull factors as well. The results also indicate that counties that are farther away from MSAs and counties located on the coast are more likely to have larger pull factors. For example, a county 100 miles away from the closest MSA will have a pull factor 40 percent higher than a county located in the MSA, all else equal.

Counties with an interstate highway tend to have higher pull factors, suggesting that the improved accessibility associated with interstate highways attracts retail business rather than facilitating its leakage out of the county. Furthermore, HIFL and HIAL show that, in fact, sales tax differentials create an incentive for consumers to cross state borders to take advantage of lower tax rates if they are near the border and have easy access to it. Thus, cross-state border effects are amplified by interstate access. This is clearly seen in the Georgia counties bordering the state of Florida, which face a lower sales tax rate than its neighboring Florida counties. Therefore, HIFL has a positive impact on Georgia county pull factors. This implies less retail sales leakages from the border county and/or capture of retail sales from border counties in Florida.

HIAL is marginally significant in the model (at the 10 percent level). In Alabama, sales tax rates are not uniform throughout the state. However, because Alabama county sales tax rates generally are lower than Georgia border counties, the effect on retail sales is negative. This supports the theory that sales tax rate differentials may induce leakage out of states with higher tax rates, in this case, the Georgia counties. In general, Model 1 yields intuitively appealing results and performs reasonably well in explaining pull factors and thereby, retail sales activity across Georgia.

As mentioned above, Model 2 does not include the variable PCI. None of the other variables show any significant changes upon the deletion of PCI, suggesting that any possible correlation between this variable and another variable--in particular DENS--is not an issue in the model.

Although Tennessee counties have a higher sales tax rate than their bordering Georgia counties, the variable is not significant in Model 3. Such results could be attributed to the joint proximity of the five Georgia counties bordering both Tennessee and Alabama. Alabama had lower sales tax rates than Georgia. Thus, the potential effects of the tax differential between Tennessee and Georgia are offset by the effects of tax differential between Alabama and Georgia. In other words, border-county Georgia consumers who faced a choice between shopping in Tennessee or Alabama may have chosen Alabama, thus obscuring the effect of the Tennessee/Georgia sales tax differential in the model. Also, the variable HISC is not significant. The sales tax rate differential between South Carolina and Georgia counties is minimal, thus the insignificance of HISC is as expected.

Model 4 in Table 1 provides results when the interactive border and highway variables are replaced by state binary variables FL, AL, SC, and TN. Not surprisingly, the state binary variables are not significant. This supports the hypothesis that tax rate differential effects are amplified and made statistically significant by easy cross-border access facilitated by interstate highways.

IMPLICATIONS

Based on the above analysis, a change in the method of financing local government away from property taxes in favor of retail sales taxes is likely to have an adverse long-term affect on the ability of local governments to fund growth in their expenditures, provided that there is no other fundamental change in the relationship between the causal variables and pull factors. The implementation of an increase in the sales tax rate could change local shopping patterns and drive sales away from Georgia counties to neighboring states. Such sales leakages could substantially hurt the revenue stream necessary to finance public programs in the state. This analysis may also apply to states like Florida, Iowa, Pennsylvania, and North Dakota that are proposing property tax reforms. Proposals that would raise the sales tax to pay for schools would result in school-funding problems because the activity on which the tax is based is more mobile for sales taxes than property taxes, in the short run.

Although in Georgia the proposed property tax reform legislation was unable to garner sufficient support for passage in the lower house, the legislature did abolish the relatively small statewide property tax levy. This action demonstrates an incremental approach with what appears to be a goal of reducing or eliminating property taxes and replacing them with an increased sales tax. The following quote (Atlanta Journal Constitution, 2007) from then-Speaker of the House in the Georgia legislature, Glenn Richardson, characterizes the strategy of some Georgia legislators:

"Instead of trying to change the entire ad valorem system in one fell swoop we'll do it a little bit at a time."

LIMITATIONS AND FUTURE RESEARCH

Several limitations of the analysis should be discussed before moving to a conclusion. The limitations also establish a potential framework for future research. The empirics herein are based on data from one year, 2006, in one state, Georgia. As such, these research findings may be subject to idiosyncratic factors influencing retail trade activity in Georgia counties in 2006, and therefore may limit the generalization of the findings to other areas and time periods. Nonetheless, the results herein are consistent with the findings of other research on retail trade activity pertaining to other states. This suggests that further research in this area may be fruitful.

As a first step, a fertile area of research may be to construct a panel data set covering multiple years for the counties of Georgia. This would allow for the consideration of broader macroeconomic conditions as they affect retail trade activity over the course of the business cycle. A cross-sectional data set is inherently unable to address this particular issue. Secondly, the panel analysis should make use of spatial analysis as in Hicks and Wilburn (2001) and Garrett and Marsh (2002) to account for potential spatial correlation across counties in the state. Ideally, the analysis could be extended in a generalized panel data set using spatial analytical techniques covering more states in the southeast or on a nationwide basis.

CONCLUSION

The focus of this paper is to explain variations in the pull factors for retail sales and to provide a framework to understand the influence of sales tax rate differentials on retail sales activity. The first hypothesis tested considered the effects of sales tax differentials on retail sales activity in border counties of Georgia. The research findings demonstrate that the combined effect of highway accessibility and border counties with the states of Florida and Alabama result in a lower pull factor for Georgia border-counties. Those border counties with a higher sales tax rate than their cross-border neighboring counties have lower retail pull factors; in contrast, counties with lower sales tax rates have higher pull factors.

A second hypothesis tested the applicability of central place theory as providing a foundation for the determinants of pull factors in Georgia counties in 2006. The results of this study confirm the findings of previous studies based on central place theory that pull factors are higher in counties with higher per capita income, more elderly population, and higher population density. In addition, pull factors are increased in counties with interstate highways, on the coast, and at greater distance from metropolitan statistical areas.

REFERENCES

American Map Corporation. (2006). United States Road Atlas. Long Island City, NY: American Map Corporation.

Atlanta Journal Constitution. (2007, November 30). Speaker to push for ending school taxes.

Broomhall, D., & King, E. (1995). Retail sales trends in Indiana counties. EC-690, Purdue University Cooperative Extension Service. West Lafayette, Indiana.

Christaller, W. (1966). Central Places in Southern Germany. Englewood Cliffs, NJ: Prentice Hill.

Federal Housing Finance Agency. (2008). MSI HPI Summary. Washington, D.C. Data retrieved 2008 from http://www.fhfa.gov/Default.aspx?Page=14.

Gale, H. F., Jr. (1996). Retail sales pull factors in U.S. counties. Review of Regional Studies, 26 (2): 177-195.

Garrett, T.A., & Marsh, T.L. (2002). The revenue impacts of cross-border lottery shopping in the presence of spatial auto correlation. Regional Science & Urban Economics, 32(4), 501-520.

Georgia Chamber of Commerce. (2007, November 27). Two months until legislature convenes; Speaker's tax reform plan to be debated. Retrieved June 27, 2010 from http://www.gachamber. com/story-7112036615AD9F 8D9.html.

Georgia Department of Revenue. 2006. Statistical Report, 2006. Atlanta, GA. State of Georgia.

Harris, T.R. (2003). Mineral County Retail Sector Leakage Study. University Center for Economic Development, University of Nevada--Reno.

Hicks, M., & Wilburn, G. (2001). The regional impact of Walmart entrance: A panel study of the retail trade sector in West Virginia. Review of Regional Studies, 31 (3): 305-13.

Hughes, D. (2004). Pull factor estimates for retail sales in West Virginia counties. Manuscript. West Virginia University Extension Service. Morgantown, West Virginia.

Iowa Department of Revenue. (2008). 423F.2 Repeal of local sales and service taxes--secure an advanced vision for education fund. Iowa Tax Research Library. Retrieved June 27, 2010, from http://itrl.idr.iowa.gov/default.asp?id=423F.2.

Kansas Department of Revenue. (2008). A Study of Retail Trade in Cities across Kansas. Office of Policy and Research. Topeka, Kansas. State of Kansas.

Mikesell, J.L. (1970). Central cities and sales tax rate differentials: the border city problem. National Tax Journal, 23 (2): 206-213.

Missouri Department of Economic Development. (2008). Missouri Retail Trade Analysis 2000-2007.

Missouri Economic Research and Information Center. Jefferson City, Missouri. State of Missouri.

Missouri Department of Economic Development. (2009). West Central Workforce Investment Area Retail Trade Analysis. Missouri Economic Research and Information Center. Jefferson City, Missouri. State of Missouri.

Nelson, M.A. (2002). Using excise taxes to finance state government: do neighboring state taxation policy and cross-border markets matter? Journal of Regional Science, 42(4): 731-752.

Norman Economic Development Coalition. (February 2009). University North Park Project Area Update. Norman, OK. Retrieved June 27, 2010 from http://www.nedcok.com/pages/home_page/news_cat/samantha_news_category_3348/sa mantha_news_story_3303/news_story.xml.

Richard Caplan and Associates. (2008). Retail Market Study: Baldwin City, Kansas. Prairie Village, Kansas.

Rohrer, S. (2009, May 9). Press release: Rohrer reintroduces school property tax elimination act. Retrieved June 27, 2010, from http://images.icuecorp.com/RI01/60043610/orgs/original/Rohrer.doc

Shaffer, R. (1989). Community Economics: Economic Structure and Change in Smaller Communities. Ames, IA: Iowa State University Press.

Shaffer, R., Deller, S., & Marcouiller, D. (2003). Community Economics: Linking Theory and Practice. Iowa State University Press: Ames, Iowa.

Stone, K.E. (1993). Impact of Walmart stores and other mass merchandisers in Iowa, 1983-2001. Manuscript. Iowa State University Department of Economics. Ames, Iowa.

Stone, K.E., & Artz, G. (2002). An analysis of the transfer of funds from weak retail counties to strong retail counties in Iowa via local option sales taxes. Manuscript. Iowa State University Department of Economics. Ames, Iowa.

Tax Foundation (2008). Florida property tax swap off ballot for good after state supreme court ruling. Tax Policy Blog. Retrieved June 27, 2010, from http://www.taxfoundation.org/blog/ show/23575.html.

Tax Foundation (2010). North Dakota may consider property tax abolition. Tax Policy Blog. Retrieved June 27, 2010, from http://www.taxfoundation.org/blog/show/26099.html.

Tosun, M.S., & Skidmore, M.L. (2007). Cross-border shopping and the sales tax: an examination of food purchases in West Virginia. B.E. Journals of Economic Analysis & Policy: Topics in Economic Analysis & Policy, 7(1): 1-20.

U.S. Census Bureau. (2006). Statistical Abstract of the United States, 2006. Washington, D.C: U.S. Government Printing Office.

Walsh, M.J., & Jones, J.D. (1988). More evidence on the 'burden tax' effect: the case of West Virginia, 1979-85. National Tax Journal, 61 (2): 261-265.

White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48(4): 817-838.

Yanagida, J.F., Johnson, B.B., Young, J., & Lundeen, M. (1991). An analysis of economic and noneconomic factors affecting retail sales leakages. Review of Regional Studies, 21(1), 53-64.

Michael Toma

Yassaman Saadatmand

Armstrong Atlantic State University

Alexis Brewer

Booz Allen Hamilton

Michael Toma (Ph.D. in Economics) is a Professor of Economics at Armstrong Atlantic State University in Savannah, GA. He also is the Director of the Center for Regional Analysis at the university.

Yassaman Saadatmand (Ph.D. in Economics) is a Professor of Economics and Head of the Department of Economics at Armstrong Atlantic State University.

Alexis Brewer (BA in Economics) earned her economics degree from Armstrong Atlantic State University in 2008. She is now working as a senior consultant for Booz Allen Hamilton's Business Analytics team in Norfolk, VA.
Table 1 Descriptive Statistics

  Variable     Maximum  Minimum   Mean   Standard Deviation

(LOG)PF        1.664   -1.711   -0.339        0.469
(LOG)PCI06    10.837    9.730   10.127        0.171
(LOG)POP65     3.254    0.588    2.444        0.322
 (LOG)DENS     7.900    2.142    4.332        1.197
 TOMSA          109       0     35.910       19.880
 COAST           1        0      0.037        0.191
 HIGH            1        0      0.289        0.455
 AL              1        0      0.101        0.302
 FL              1        0      0.063        0.244
 SC              1        0      0.082        0.275
 TN              1        0      0.044        0.206
HIAL             1        0      0.033        0.175
HIFL             1        0      0.013        0.111
HISC             1        0      0.013        0.111
HITN             1        0      0.013        0.111

Table 2
Correlation Matrix

             (LOG)PF   (LOG)PCI   (LOG)POP65   (LOG)DENS

LOG)PF        1.00        --          --          --
(LOG)PCI06    0.51       1.00         --          --
(LOG)POP65    0.01      -0.26        1.00         --
(LOG)DENS     0.52       0.68       -0.47        1.00
HIGH          0.35       0.36       -0.30        0.46
COAST         0.14       0.15       -0.20       -0.06
TOMSA        -0.16      -0.46        0.37       -0.37

             HIGH    COAST   TOMSA

(LOG)PF       --       --      --
(LOG)PCI06    --       --      --
(LOG)POP65    --       --      --
(LOG)DENS     --       --      --
HIGH        1.00       --      --
COAST       0.31      1.00     --
TOMSA      -0.30     -0.24    1.00

Table 3
Regression Results: Dependent Variable is (log) PF.

  Variable          Model 1           Model 2

C            -10.14 (-3.99)        -3.00 (-10.33)
LOG(PCI)        0.74 (2.88) **           --
LOG(POP65)      0.49 (5.89) **        0.54 (7.06) **
LOG(DEN)        0.20 (6.27) **        0.27 (9.57) **
HIGH            0.13 (2.18) **        0.13 (2.06) **
COAST           0.28 (3.03) **        0.35 (4.54) **
TOMSA          0.004 (2.51) **       0.002 (1.72)
HIFL            0.48 (7.26) **        0.46 (9.71) **
HIAL          -02.20 (-1.61)         -0.16 (-1.96) *
HISC                 --                  --
HITN                 --                  --
FL                   --                  --
AL                   --                  --
SC                   --                  --
TN                   --                  --
[R.sup.2]           0.45                0.42
Adj[R.sup.2]        0.42                0.39
F                  15.82                15.75
n                   159                  159

  Variable          Model 3           Model 4

C              -10.15 (-3.97)      -10.03 (-3.74)
LOG(PCI)          0.74 (2.86) **      0.72 (2.69) **
LOG(POP65)        0.50 (5.84) **      0.48 (5.18) **
LOG(DEN)          0.20 (6.24) **      0.21 (6.04) **
HIGH              0.13 (2.14) **      0.12 (1.98) **
COAST             0.29 (3.31) **      0.33 (3.12) **
TOMSA            0.004 (2.46) **     0.004 (2.29) **
HIFL              0.47 (7.06) **          --
HIAL            -0.21 (-1.61)             --
HISC            -0.08 (-0.64)             --
HITN             -0.01 (0.08)             --
FL                    --              0.34 (1.83)
AL                    --              0.03 (0.26)
SC                    --              0.01 (0.23)
TN                    --             -0.00 (-0.04)
[R.sup.2]            0.45                0.27
Adj[R.sup.2]         0.42                0.43
F                   12.50                13.28
n                    159                  159

Numbers in parenthesis are t-statistics

** Significant at the 1 percent level.

* Significant at the 5 percent level.
COPYRIGHT 2010 International Academy of Business and Public Administration Disciplines
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2010 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Toma, Michael; Saadatmand, Yassaman; Brewer, Alexis
Publication:International Journal of Business and Economics Perspectives (IJBEP)
Article Type:Report
Geographic Code:1U5GA
Date:Jun 22, 2010
Words:5969
Previous Article:Religion's impact on corruption.
Next Article:Korean and American student earnings estimates.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters