The Equity of Local Sales Tax Distributions in Urban,Suburban,Rural,and Tourism Rich Counties in North Carolina

Abstract

The use of local options sales taxes (LOSTs) is creating largely unexplored equity concerns with regard to the revenue raising capabilities of different local governments. This article focuses on differences among urban, suburban, and rural counties; the impact of proximity to urban (or retail) centers; and the impact of LOST decisions on tourism rich counties, a new category of local governments. Using data from 2003 to 2009 on all 100 North Carolina counties and a spatial Durbin error panel model, I identify factors relating to LOST revenue raising capacity (RRC). The results indicate that tourism rich counties have the greatest LOST capacity, suburban counties have the least, and there is a penalty for bordering an urban county. However, statistically significant differences in the RRC of the four types of counties disappear once property taxes are included in the mix.

Keywords

local sales taxes urban tourism revenue raising capacity

There is a presumption in the literature on local option sales taxes (LOSTs) that urban jurisdictions benefit from LOSTs to the detriment of the nonurban jurisdictions that subsidize them with their citizens’ tax dollars. This study examines the relationship between the urban designation and the equity of LOST revenue raising capacity (RRC), as well as overall RRC, in North Carolina (N.C.) counties from 2003 to 2009. Using a four-step approach that tests the assumptions of previous research and advanced methodologies, including spatial Durbin error panel model (SDEM), this article finds that LOST RRC is inequitable. Urban counties generate almost US$6 per capita more from LOSTs than do suburban counties and tourism rich counties generate over US$13 per capita more than suburban counties. However, an examination of overall RRC (LOSTs and property taxes) reveals no statistically significant difference between urban classifications, suggesting that LOSTs may complement property taxes and improve the equity of overall RRC.

LOSTs have proliferated over the past thirty years. Within the thirty-seven states that currently permit their adoption, more than 11,000 local governments have one in place, and it is their second largest source of own-source revenue (Tax Policy Center 2006; Drenkard 2013). This widespread use and growing reliance on LOST revenue has raised concerns about RRC inequities. Research that shows urban areas and regional retail centers are more likely to institute a LOST, presumably because they can transfer the burden to nonresidents (Burge and Rogers 2011; Burge and Piper 2012), echo this concern. Varying abilities to export the burden are perceived as creating or exacerbating inequities between local governments. However, little previous research has examined the validity of these perceptions directly. To address the perceived revenue raising disadvantage of rural jurisdictions, N.C. is considering replacing LOSTs with a state sales tax that would be redistributed on a per capita basis amongst local governments.

Zhao and Hou’s (2008) study of the relationship between urban and rural counties uses descriptive statistics to examine the equity of distributions. Their study is foundational to this analysis and concludes that LOSTs do not “exacerbate fiscal inequality” (p. 54). The present study builds on that work by empirically analyzing both the LOST revenues generated and the effect of LOST revenues on overall RRC while controlling for many outside factors, including urban classification. This study finds that LOST capacity varies among classifications, but the differences in overall RRC between counties with different urban classifications are not statistically significant when property taxes are included. In addition, this study goes beyond classifying counties as urban or other by expanding the classifications to include suburban and tourism rich. This differentiation is critical to better understanding which jurisdictions benefit most from tax importation. Finally, this article addresses the possibility of a spatial component of tax exportation by controlling for adjacency of urban counties and finds that counties with urban neighbors generate less LOST revenue than do counties without urban neighbors.

My methodology tests previous findings and assumptions through a four-step process. The first step examines whether urban counties generate more LOST revenue per capita than their rural and suburban counterparts and finds that urban counties generate more revenue per capita than suburban counties but not more than rural counties. In the second step, an SDEM controls for spatial autocorrelation and dependence and finds a statistically significant positive relationship between being an urban county and LOST RRC and a negative relationship between proximity to an urban county and LOST RRC. The third step introduces an additional category of county: tourism rich. As expected, tourism rich counties generate more revenue from LOSTs than do their urban, suburban, and rural counterparts. The fourth and final step tests whether this disparity in LOST RRC results in greater inequities in overall RRC from major own-source revenues (LOSTs and property taxes). The results suggest that the urbanism of a county does not affect overall RRC. Coupled with the results from the previous step, this finding suggests that LOST revenue may actually improve RRC equity. This finding adds to the literature that has suggested that property tax base and sales tax base may be uncorrelated or even negatively correlated, perhaps leading to equity gains when LOSTs are added to local government tax portfolios (Ghaus 1995; Craft 2002; Zhao and Hou 2008; Wang and Zhao 2011). These findings are a valuable addition to the ongoing conversation at both the local and state level about equity and capacity of local governments and should be used to inform the decision making and expectations of counties considering the adoption of LOSTs.

I organize this article as follows. The second section describes the LOST laws of N.C. and the institutional context of the study. The third section reviews pertinent literature on LOSTs and where this study fits into the existing literature. The fourth section outlines the strategy for identifying the winners and losers of LOST use. The fifth section describes the data and estimation methodology: descriptive statistics, a fixed effects model, and SDEM. The sixth section discusses the results of the empirical analysis. The seventh and concluding section discusses the implications.

Overview of LOSTs and Select County Financing Considerations in N.C.

N.C. has many characteristics that make it an ideal choice for analyzing the relationship between LOST RRC and urbanism. Because all 100 county governments have implemented a LOST, many restrictions and concerns noted in previous studies do not pertain. There is not a self-selection or exclusion problem, nor are there vertical tax differentials of concern. In N.C., only county governments can adopt a LOST, not municipalities. This is important because previous studies have found that when both counties and municipalities have the ability to adopt LOSTs, the decisions of one level of government in the vertical affect the other (Burge and Piper 2012). Additionally, because there is almost no diversity in the LOST rate among N.C. counties and there are no municipal rates, horizontal tax differentials also are of little concern. This is important because research has shown a strong relationship between tax competition and rate setting (Burge and Piper 2012). Finally, it is possible in N.C. to identify the exact size of the tax base due to the ability to isolate the revenue generated by each LOST instrument, the rates of which do not vary.

In the analysis period 2003 to 2009, the state allowed counties to levy four separate local sales and use taxes (outlined as articles in the N.C. General Statutes) with a combined rate of 2.5 percent. During this period, all 100 counties had all four articles in place (N.C. Department of Revenue 2013a). Article 39, the first LOST levied in N.C., became effective in 1971 and has a rate of 1 percent. It is distributed on a point-of-sale basis.¹ Article 39 was followed by Articles 40 and 42, which were enacted in response to a state reduction in funding for county-level primary and secondary education capital funding. As such, they are partially earmarked for capital spending on education. They each have a rate of 0.5 percent and are unusual on two fronts: they are distributed to counties on a per capita basis, not point-of-sale, and they are both partially earmarked. Article 44, which had a rate of 0.5 percent,² was repealed in 2009. Fifty percent of the revenue generated by Article 44 was returned to the county in which the goods were sold and the other 50 percent was distributed by the weighted per capita method similar to that used for Articles 40 and 42.

While N.C. provides a unique institutional context³ to examine the RRC of LOSTs, it is also representative and comparable to other states in many important ways, such as its diversity among counties. The counties fall along all parts of the wealth spectrum, from urban, affluent counties to poor rural counties. Many areas, mostly in the mountains and on the coast, have a booming tourism industry.⁴ Also N.C. counties are similar to the national average in two key demographic measurements: (1) the average population of the state’s counties is 95,602 compared to 98,418 nationally and (2) the median family income among N.C. counties is US$52,404 compared to US$50,636 nationally. N.C. counties do spend significantly less on services than their interstate peers: US$625 million in N.C. versus the national average of US$1.1 billion.⁵ The results of this study should provide insights nationally.

Literature Review

Goals for LOST Use

The LOST is one of the more popular revenue sources available to local governments because LOSTs have relatively low visibility and typically require voter approval (Advisory Commission on Intergovernmental Relations 1986; Oates 1991; Schwartz 1997; Wachs 2003). Four primary goals have been attributed to the use of this revenue source, explaining its widespread appeal: (1) increase revenue and fund additional services, (2) provide property tax relief, (3) increase revenue stability through diversification, and (4) export a portion of the tax burden to noncitizens (Jung 2001). The first three reasons are relatively self-explanatory and have been explored deeply in the literature (see, e.g., Sjoquist, Walker, and Wallace 2005; Afonso 2013, 2014; Carroll 2009; Hou and Seligman 2008). The fourth, tax exportation, is the most pertinent to this study.

Local governments may adopt LOSTs because they believe the generated tax dollars will come from nonresidents. “Tax exportation occurs when the economic incidence of a tax levied by one jurisdiction is at least partially shifted to individuals residing outside the tax jurisdiction” (Burge and Piper 2012, 394). One of the primary characteristics of early adopters is whether the local government anticipates being able to export its burden (Zhao 2005; Sjoquist et al. 2007; Burge and Piper 2012). However, not all local governments are able to import tax dollars—some become exporters. Tax exportation is central to this study because it is a primary reason why urban governments collect more revenue than nonurban governments and makes the examination of tourism rich areas necessary.

Types of Governments That Use LOSTs

It seems reasonable, given limitations restricting local government fiscal autonomy, that one of the primary goals of adopting a LOST is to increase RRC, even if it is accompanied by a reduction in property tax burdens. However, previous studies suggest that a local government’s ability to generate substantial LOST revenue is highly dependent on whether it is an urban or a regional retail center that will benefit from other communities’ tax exportation (Rogers 2004; Zhao and Hou 2008; Burge and Rogers 2011; Burge and Piper 2012). “Retail rich” communities are jurisdictions with a tax surplus. “Retail poor” communities are those that export tax dollars to the “retail-rich” communities (Artz and Stone 2003). Typically, researchers use total dollars generated, which explains why the vast majority of research has shown that urban communities fare so much better than nonurban communities (Rogers 2004; Cornia et al. 2010; Burge and Rogers 2011; Burge and Piper 2012).⁶ However, per capita measurement has shown that urban counties generate more LOST revenue than their nonurban counterparts (Zhao and Hou 2008).⁷

Due to differences in the size of rural and urban tax bases, the literature appears united in the assessment that an uneven distribution of LOST revenue exists among local governments. This is why many consider LOSTs inequitable. However, if a local government’s sales tax base is either uncorrelated or negatively correlated with the tax base of other revenue sources (e.g., the property tax), a LOST may not add to the overall disparity and may actually improve equity (Craft 2002; Zhao and Hou 2008; Wang and Zhao 2011).

Identification of LOST Winners and Losers

I define winners and losers, with respect to LOST revenue, by whether the county becomes an importer or an exporter of sales tax dollars after adoption of a LOST. Losers experience “tax leakage” (Artz and Stone 2003, 2). Winners are those local governments that not only capture sales tax revenue from its own citizens but also import tax dollars from nonresidents. Losers do not benefit from importing nonresident sales tax dollars; instead, they export (i.e., leak) LOST revenue to other jurisdictions.⁸ This study classifies winners and losers by identifying statistically significant differences in the per capita LOST revenue receipts across counties with different urban classifications. After controlling for factors such as the wealth of the county and the unemployment rate, the classifications with the highest LOST revenues per capita are classified as the winners (i.e., they benefit from tax importation) while those with the lowest LOST revenues per capita are classified as the losers (i.e., they export tax revenues).

This analysis uses a framework⁹ similar to that of Burge and Piper (2012), which assumes the following:

Every county has a LOST in place throughout the entire period under analysis.

County governments must balance their budgets.

County governments do not compete for LOST dollars through their LOST rates (i.e., the tax differentials within the state are inconsequential).

Consumers prefer not to travel long distances to make their purchases.

Consumers prefer retail centers in clusters, with numerous retail options available within each center.

The landscape of N.C. counties is consistent with these assumptions and therefore makes the state particularly well suited for this analysis. Assumption 3 is met because unlike such states as Oklahoma where rates vary from 0 percent to 5 percent, for 686 of 700 observations in N.C. during the period of study the overall rate was 2.5 percent. The exceptions were six counties that levied an additional 0.25 percent rate starting in April 2008, one county that had an additional 1 percent rate for six months in 2006, and one county that levied an additional 0.5 percent rate for public transportation for the entire period (N.C. Department of Revenue 2013a).¹⁰ The four steps of analysis in this study are described subsequently.

The first step examines whether urban jurisdictions generate more LOST revenue per capita than their nonurban counterparts, as reported in the literature (Rogers 2004; Zhao and Hou 2008; Burge and Rogers 2011; Burge and Piper 2012). In contrast to previous studies, I prefer to examine LOSTs in per capita terms because, by taking into account how many residents constitute the tax base, I capture tax leakage more accurately. I expect a county that has triple the population of its neighbor to generate more LOST revenue in absolute terms whether or not it imports additional revenue. I expect to find that urban counties will experience tax surpluses (i.e., be winners) and that rural communities will experience tax leakage (i.e., be losers).

The second examines the possibility of regional effects. Consumers may choose to shop outside of their jurisdiction because of preferences for retail agglomerations, but this is tempered by a preference for shorter driving times (Burge and Rogers 2011). Therefore, consumers are more likely to shop in neighboring jurisdictions than across the state. I expect to find that counties with urban neighbors experience more tax leakage than those that are more isolated. This is in keeping with assumptions 4 and 5. The regional effects capture in-state importing and exporting of LOST revenue.

In the previous steps, I classify counties as urban, suburban, or rural. In step 3, I introduce a fourth county designation: tourism rich. N.C. has both a coastal region and a mountain region that attract a great deal of tourism from state residents and nonresidents alike. I expect to find that these communities, though perhaps demographically similar to suburban or rural counties, will benefit greatly from tax importation due to the increased number of nonresident visitors (i.e., tourists) and, therefore, deserve special consideration. The tourism rich classification captures both in-state and out-of-state importing of LOST revenue.

Fourth, after determining the winners and losers of LOSTs in the prior three steps, I examine the overall winners and losers of the two revenue streams: property taxes and LOSTs. If property taxes and LOSTs are substitutes for one another (Ghaus 1995), counties characterized as losers with regard to LOST revenue may be winners with regard to property tax revenue. I expect the four types of counties identified (urban, suburban, tourism rich, and rural) will be distinct and that the effect of LOSTs on overall RRC will be offset by action taken to tap the capacity available in property taxes. Overall, I expect the biggest losers will be rural counties because they have both a smaller sales tax base and property tax base. The biggest winners will be tourism rich counties in terms of both LOST revenue and property taxes. I expect an inverse relationship between suburban and urban counties: suburban counties will have a strong property tax base, whereas urban areas will have a strong LOST base. Suburban jurisdictions often are bedroom communities having a disproportionately residential tax base, and urban jurisdictions are more likely to benefit from incoming commuters, tourists, and retail agglomerations.

Data and Methodology

I study the relationship between urbanism and LOSTs through examination of descriptive data and an empirical analysis using a fixed effects model and SDEMs using data from all 100 N.C. counties over the period 2003 to 2009. This section provides descriptive data and empirical strategies for each of the four steps.

Step 1—Revenue Generation

Descriptive data

To see if urban counties actually generate more LOST revenue, I look at both total dollars (in keeping with the previous literature) and per capita LOST collections (to examine overall equity) by urban, suburban, and rural (i.e., nonurban) counties in N.C. For this study, a county is classified as urban if it has a population greater than 200,000.¹¹ Population data are drawn from the Bureau of Economic Analysis (BEA; 2015), which I use when constructing the denominator for per capita measures. I use the National Center for Health Statistics (NCHS) designation of large fringe as suburban. The NCHS characterizes these fringe counties as suburban counties with large central metros that have populations of at least one million. Other counties are considered “neither” and are therefore categorized as rural.

The LOST revenue distributions come from the N.C. Department of Revenue (2013a). In this analysis, I use only LOST revenue generated by Article 39 because, unlike the other LOSTs in N.C., its revenue is distributed exclusively by point-of-sale. This is the most typical way that LOST revenue is distributed—and the most applicable outside N.C.¹² I consider the revenue in both total and per capita dollars.

Table 1 reports descriptive statistics and demonstrates, as previously suggested in the literature, that urban counties generate much larger sums of money than their nonurban counterparts. Urban counties generate an average of US$41 million in LOST receipts compared to less than US$9 million for suburban and US$4 million for rural counties. However, the table also tells another story. Urban counties, on average, bring in more money than their rural and suburban counterparts on a per capita basis, but only by a small margin. Urban counties have an average of US$89.75 per capita in LOST revenue compared to US$84.74 and US$68.36 for suburban and rural counties, respectively. It is reasonable to expect that the per capita measures would make the revenue distributions appear more equitable; however, because of retail agglomerations and tax exportation, the conventional wisdom is that urban areas will still enjoy a sizable RRC advantage. Table 1 emphasizes the importance of steps 2 and 3 in this analysis. Although urban counties, on average, generate the most per capita LOST revenue, a different picture emerges at the extremes among rural counties. Rural counties have the lowest and the highest per capita collections by far. Perhaps an important distinction among these rural counties is tourism.

Table 1.

LOST Revenue by Urban Classification.

		Mean	Minimum	Maximum
In total dollars	Urban	41,300,000.00	10,100,000.00	109,000,000.00
	Suburban	8,728,463.00	917,598.00	16,900,000.00
	Rural	4,094,665.00	161,492.00	26,800,000.00
	State average	6,986,206.00	161,492.00	109,000,000.00
In per capita dollars	Urban	89.75	47.16	126.26
	Suburban	84.74	35.57	168.46
	Rural	68.36	16.12	295.27
	State average	70.71	16.12	295.27

Note. Tourism rich counties are classified above as either suburban or rural. LOST = local options sales tax.

Another way of examining the data is to look at the spatial distributions of the data. Figure 1 presents a visual representation of N.C.’s counties with regard to population and LOST revenue (Article 39 only) both in total revenue and per capita terms.¹³ The population and total LOST revenue maps are similar, as expected, reinforcing the idea that the counties that generate the most LOST revenue are the most populous.

Figure 1.

North Carolina’s population, local options sales tax (LOST) revenue, and LOST revenue per capita.

However, as suggested by table 1, the map of LOST revenue per capita deviates quite a bit from the population map. Suburban counties, located on the central southern border of the state around Mecklenburg County (Charlotte), are generating approximately the same amount of revenue per capita as urban areas even though in total dollars they generate about a fifth as much on average. Furthermore, some of the least populous counties are generating the most revenue per capita (see table 2). This is especially true for the western and eastern portions of the state, where the mountain and coastal counties are located, respectively.

Table 2.

Average Property Tax, LOST Revenue, and Revenue Raising Capacity.

		LOST	Property Tax Revenue	Total Property Value	Revenue Raising Capacity
In total dollars	Urban	14,700.00	291,709.00	44,000,000.00	315,000.00
	Suburban	5,463.00	41,237.00	8,225,920.00	58,800.00
	Rural	3,020.00	19,948.00	3,308,751.00	31,900.00
	Tourism rich	8,953.00	60,061.00	12,300,000.00	77,600.00
In per capita dollars	Urban	89.75	610.00	93,316.00	685.48
	Suburban	89.53	633.00	143,450.60	1,019.72
	Rural	64.54	465.00	79,798.64	594.49
	Tourism rich	100.71	554.00	132,354.30	969.18

Note. The suburban and rural counties that have been recoded as tourism rich are not included in the suburban and rural averages. LOST revenues, property tax revenues, total property values, and revenue raising capacity (RRC) values in total dollars are presented in thousands of dollars; the per capita terms are not. LOST = local options sales tax.

Regression strategy

I hypothesize that urban counties will generate more revenue per capita than their suburban and rural counterparts. However, the descriptive statistics suggest that the margin may be smaller than expected. The initial specification of the model will use the three county classifications: urban, suburban, and rural. To test this hypothesis, I model the relationship using a fixed effects model with the standard errors clustered by county. The dependent variable is per capita sales tax base, as previously described. The independent variables are the millage rate, per capita nonproperty tax and nonsales tax revenue, median family income, total income of county residents, binary variables for urban and rural, percentage of the population that is over age sixty-five, population density, and unemployment rate. Fixed effects remove the effect of the unobserved time-invariant characteristics that are specific to individual counties. I perform a Hausman (1978) test using the fixed effects model. The estimated χ² is 29.23 rejecting the null hypothesis that the difference between the coefficients estimated under random effects and fixed effects is not systematic.

The millage rates are not available for the entire period, so I calculate them by taking the county’s property tax levy and dividing it by the total assessed valuation. The per capita nonproperty tax and nonsales tax revenue constitute the total revenue receipts for the county minus property tax revenue and total LOST revenue. I take the levy, the assessed valuation data, and total revenue data from the N.C. Department of Revenue (2013b). I collect the estimated median family income from the Department of Housing and Urban Development (2010). I take the total income of county residents from the BEA (2015). The two income variables are included to control for the effect of the county’s wealth on its ability to generate LOST revenue. The three categories used in step 1 for characterizing the urbanism of the county are urban, suburban, and rural, as previously described.

I include the population density of the respective county because an analysis of rural areas that are becoming increasingly populated finds that the more an area is geographically dispersed, the higher its costs of providing services (Webb and Anderson 2005). This implies that controlling for density in addition to urbanism may be valuable. Population density and percentage of the population that is over sixty-five are from the N.C. Office of the Governor (2013). The unemployment rate is taken from the N.C. Department of Commerce (2013b).

Step 2—Regional Considerations

Descriptive data

When examining the winners and losers of LOSTs, it is important to consider the possibility of a spatial component, meaning that there may be a spillover effect from neighboring counties that affects LOST revenue collections. I do this in two ways: local Moran’s I and global Moran’s I. A local Moran’s I, often referred to as a local indicator of spatial autocorrelation (LISA), identifies which observational units within the data set are partially autocorrelated with their neighbors (Anselin 1995).¹⁴ Products of this analysis appear in an online appendix.

LISA also helps identify hotspots of spatial autocorrelation. Global measures may not be sufficient because the spatial relationships that may exist in one region may not be the same elsewhere. For example, two rural counties that are adjacent could both have low LOST revenues (low–low) and, therefore, positive spatial autocorrelation, whereas a rural county that neighbors an urban county could have negative spatial autocorrelation because the rural county might have low LOST revenue while the urban county might have high revenues (low–high). The LISA maps also demonstrate that the spatial autocorrelation is very different when looking at LOST revenue (clustered in the northeast of the state and in the south central portion of the state) and LOST revenue per capita (clustered in the northeast of the state, the western portion of the state, and in a few counties in the center of the state). This is not surprising in light of the data presented in table 1.

The local spatial autocorrelation is also interesting when taking into consideration the urban classification of these counties. For LOST revenue, the cluster of high–high counties in the south central region of the state encircles Mecklenburg County, where Charlotte is located. I expect substantial LOST revenue would be generated in that region. Similarly, in the northeast region of the state is a cluster of low–low counties. This is an extremely rural area, so it is also not surprising that very little LOST revenue would be collected there.

In contrast, when examining the local spatial autocorrelation of LOST revenue per capita, a different pattern emerges. The high–low counties are scattered throughout the state, demonstrating the prevalence of urban and rural neighbors in N.C. However, what surfaces are two clusters of high–high counties: one in the far west and one along the northeast coast. This also is expected. These are the mountain and the Outer Banks regions, both areas with low population density but with rich tourism industries. These patterns reinforce the need to consider tourism rich as a fourth classification.

Researchers rarely use the LISA estimates for inference, but instead use the (global) Moran’s I, the most frequently used measure of global spatial autocorrelation. The public finance literature uses Moran’s I to measure such spatial autocorrelation topics as excise tax revenue, redistribution policy, and state-level tax policy (Rork 2000; Tosun 2003; Minkoff 2009). The following equation represents its calculation (Moran 1948). The i and j are the index of the counties, w_ij is the weights matrix, where w_ij equals zero for any i and j combination that are not neighbors (i.e., when calculating the covariance term between counties, it is considered only if the counties are identified as neighbors).

I = [\frac{n}{\sum_{i = 1}^{n} \sum_{j = 1}^{n} w_{i j}}] \frac{\sum_{i = 1}^{n} \sum_{j = 1}^{n} w_{i j} (y_{i} - \overset{ˉ}{y}) (y_{j} - \overset{ˉ}{y})}{\sum_{i = 1}^{n} {(y_{i} - \overset{ˉ}{y})}^{2}} .

For 2003, the initial year in this analysis, Moran’s I for LOST revenue per capita is 0.294, with a standard deviation of 0.059. The Moran’s I value will range from −1 to 1, where 0 represents no spatial autocorrelation, −1 represents perfect negative spatial autocorrelation, and 1 represents perfect positive autocorrelation. A value of 0.294 can be interpreted as moderate positive spatial autocorrelation between contiguous neighboring counties. Positive spatial correlation suggests that neighbors are alike, so counties with high LOST revenue are likely to have neighbors with high LOST revenue.

Regression strategy

I hypothesize a spatial component to the relationships between urban and nonurban counties in amount of revenue they raise from LOSTs because of the preference for greater or more diverse retail options and the sometimes conflicting preference for closer retail options. The calculated Moran’s I suggests positive spatial autocorrelation, and there are concerns of both spatial heterogeneity and spatial dependence. Spatial heterogeneity is not a spatial interaction, but spatial structural effects, such as environmental considerations (i.e., on the coast or in the mountains). Spatial dependence is of more interest to this study and can be understood as what happens in one area impacts that area’s neighbors and vice versa. Spatial dependence occurs when there is interaction among social and economic actors. An example of this spatial dependence occurs when the citizens of County A cross the border into County B to spend money and thereby generate LOST revenue for County B.

I address the issue of spatial autocorrelation using an SDEM with fixed effects for county and year. SDEM is chosen because it accounts for both the possibility that an unobservable variable is spatially correlated (e.g., geographic characteristics) and the possibility that there are exogenous spatial interactions among the independent variables (e.g., median income; Golgher and Voss 2014). The model is represented as follows:

y_{t} = C_{t} {\hat{δ}}_{1} + U_{t} {\hat{β}}_{1} + W I N C_{t} {\hat{δ}}_{2} + W U R B_{t} {\hat{β}}_{2} + ϵ_{i} + τ_{t} + u_{t},

u_{i t} = μ_{i} + v_{i t},

with t = 1, 2, … , T and i = 1, 2,…, N.

In this model,¹⁵ y_t is an (N × 1) vector of LOST revenue per capita, STB_it . C_t is a (N × k) matrix of county characteristics (millage rate, median family income, percentage of the population that is over sixty-five, population density, and unemployment rate), and U_t is a (N × g) matrix of measures capturing the categories expressed previously (urban and rural). W is a (N × N) spatial weights matrix. The weights matrix is a first-order row-standardized queen contiguity matrix and is implemented using the Stata module USSWM (Merryman 2008).¹⁶ The spatial lags are simply the average value of all of the neighboring counties (w_ij is equal to zero unless county i and j are neighbors).¹⁷ I include WINC _t , the spatially lagged median family income measure representing the weighted average of neighboring counties’ median incomes. This controls for the possibility of tax importation to capture the possibility that with wealthier neighbors, the county may be able to attract more LOST dollars. WURB _t is the spatially lagged urban measure and identifies the percentage of neighboring counties categorized as urban.¹⁸ I include it because the literature suggests that people prefer shorter distances and retail agglomerations; thus, counties with urban neighbors may export more LOST dollars. In addition, previous literature has shown that rural communities that are not near urban centers have a greater capacity to be retail centers and generate more revenue (Rogers 2004; Zhao and Hou 2008). ${\hat{δ}}_{2}$ and ${\hat{β}}_{2}$ are the spillover effects from neighboring counties on these two characteristics. $ϵ_{i}$ and τ_t are county and year fixed effects. u_t is a (N × 1) vector of error component disturbances. These spatial lags capture a portion of the importing/exporting of LOST revenue. A Hausman (1978) test is performed using the SDEM. The estimated χ² is 64.31 rejecting the null hypothesis that the difference between the coefficients estimated under random effects and fixed effects is not systematic.

Step 3—Tourism Rich

Descriptive data

The typical characterization of counties or municipalities is urban and other.¹⁹ In steps 1 and 2, I have also designated a third category: suburban. Suburban areas lie on the fringes of urban centers, meaning their residents have a relatively short distance to travel for many shopping options. Theory suggests that this makes suburban residents more likely to shop outside of their county (Abdel-Rahman 1990; Burge and Rogers 2011).²⁰ I create an additional category: tourism rich, nonurban areas that nonresidents regularly visit, bringing their sales (and other) tax dollars.

I define an area as tourism rich using data from the Travel Economic Impact Model provided by the N.C. Department of Commerce (2013a). Its measure of total traveler expenditures by county is the basis of the tourism measure. I classify an area as tourism rich if more than US$150 million dollars is spent on traveler expenditures within the county in one fiscal year and if the county is not characterized as urban.²¹ The threshold of US$150 million includes the top 15 percent of rural and suburban counties in terms of tourism expenditures.²² I exclude urban counties, but reclassify two suburban counties and ten rural counties as tourism rich. Of these ten counties, I classify only seven as tourism rich for the entire seven-year period.

Table 2 confirms the value of considering tourism rich counties separately. It uses the same data as table 1, presenting LOST revenue in total dollars and in per capita terms, but recategorizes some suburban and rural counties as tourism rich. This new distinction shows that tourism rich areas generate a great deal in total dollars and more than their urban counterparts on a per capita basis. Urban areas generate the most revenue from Article 39, and the difference in the average LOST per capita between rural and tourism rich is significant, as expected.

Regression strategy

I hypothesize that tourism rich counties are different from their suburban or rural peers because, theoretically, they import LOST revenues at a higher rate. For step 3, I test this by reclassifying the counties based on the tourist dollars spent in the county. To test the relationship, I use the same specification used in step 2 but with the updated suburban and rural measures and the tourism rich classification for U_t . The tourism rich classification captures a portion of the ability to import LOST revenue from nonresidents, both from within and outside of the state.

Step 4—RRC

Descriptive data

Finally, I consider property tax revenue in addition to LOSTs in order to measure the overall winners of N.C. county revenue structure in terms of RRC.²³ Some suggest that LOSTs are less inequitable than they appear to be because they may be negatively correlated with property taxes (Craft 2002). As laid out in the fourth section, I expect suburban and tourism rich areas to be the winners in the property tax revenue distributions and rural and urban areas to be the losers. I identify property tax winners and losers using two types of property tax variables: ad valorem taxes and total assessed property valuation. Data on actual ad valorem taxes come from the N.C. Department of the State Treasurer (2013). The N.C. Department of Revenue (2013b) provides data on total assessed property valuation.

Table 2 suggests that tourism rich areas are winners of both LOST and property taxes, as hypothesized. tourism rich counties have the highest LOST revenues per capita. While their per capita property tax revenue is lower than that of suburban or urban areas, tourism rich areas appear to have excellent RRC and are simply able to keep property tax burdens lower than those of their peers. In contrast, rural areas are the clear losers for both LOST and property tax revenues, also in keeping with my hypotheses. Non-tourism rich rural counties have the least capacity for both property taxes and LOSTs.

The story becomes slightly more complicated when looking at suburban and urban counties. Suburban counties generate more property tax revenue per capita than any of the other types of counties, and they have the most capacity per capita. As for LOST revenue, although the total dollars are relatively low, the per capita measure is middle of the pack. Suburban counties are arguably neither winners nor losers in the contest for LOST revenue, but they perform better than expected.

Urban counties do not behave as previously assumed; in total dollars, they are the winners with regard to both LOST and property tax revenues, but when converted to per capita terms, they are only average for LOST revenue per capita. They may generate less LOST revenue per capita than anticipated, but they also generate more property tax revenue per capita than expected. Urban counties are only behind suburban counties for property tax revenue per capita but are lower on the property valuation per capita than suburban or tourism rich areas, suggesting they have higher millage rates.

Regression strategy

To test the overall winners and losers of LOST and property tax revenues, step 4 examines the data with a new dependent variable, that is, RRC. As table 1 suggests, there is a wide range in both the total and per capita dollars collected by LOSTs and property taxes. This specification of the model measures how the different types of counties fare with regard to both LOSTs and property taxes.

Using the representative tax system strategy for capturing differences in the RRC of counties, I examine if there are differences based on the urban classification of the county. This is one of the two methods used by Zhao and Hou (2008) to examine fiscal disparities in Georgia counties. Their measure of RRC, RRC_i , is calculated by taking the sum of all the tax bases in the state, BASE_ik , and weighting them by the representative tax rates, TAX_ik , where i is the notation for county, t is for year, and k is for the tax instrument.

R R C_{i} = \sum_{k} T A X_{i k} B A S E_{i k} .

As laid out in the fourth section, this study focuses on the two major taxes available to counties, property taxes and LOSTs, consistent with Zhao and Hou’s (2008) approach. Thus, RRC_i is the sum of the valuation of the county’s per capita property, PTB_it , weighted by the average millage rate in the state for the year, MIL_t , and the per capita sales tax base, STB_it , weighted by the average LOST rate, LOST_t . In this analysis, I use only LOST revenue generated by Article 39, which has a uniform 1 percent rate across the state (i.e., LOST_t is 1 percent for the entire period).

R R C_{i t} = P T B_{i t} \times M I L_{t} + S T B_{i t} \times L O S T_{t} .

PTB_it is the total assessed property valuation in the county divided by the county’s population. I create the average millage rate in the state for the year, MIL_t , by averaging the calculated county millage rates in the state each year. The importance of using the average millage rate and the property tax base instead of just the per capita property tax revenue generated is that this measure seeks to identify the RRC, that is, not simply what the counties are raising, but what they could be raising—removing the policy choice. I use an SDEM to test these relationships, specifically, an SDEM with county and year fixed effects and the same controls as step 3, with the exception of the millage rate, which I no longer include as an independent variable. I present the summary statistics of these in table 3.

Table 3.

Summary Statistics.

Variable	Observations	Mean	Standard Deviation	Minimum	Maximum
RRC	700	654.062	398.169	250.216	3,636.664
Other Revenue	700	0.652	0.376	0	3.360
Family Income	700	49.094	8.514	32.9	76.9
Total Income	700	2,960.000	5,260.000	84.300	41,200.000
Urban	700	0.076	0.264	0	1
Suburban	700	0.050	0.218	0	1
Rural	700	0.793	0.406	0	1
Tourism Rich	700	0.106	0.308	0	1
Over 65	700	14.524	3.710	6.842	25.522
Unemployed	700	3.193	1.141	1.236	7.483
Density	700	191.587	250.455	8.400	1,683.900
LOST pc	700	70.707	32.832	16.120	295.266
Mills	700	6.513	1.663	1.875	12.892
Property Value	700	89,855.340	58,225.860	30,597.040	520,826.200
LOST	700	6,986,206	12,400,000	161,492	109,000,000
Variable	Description
RRC	Revenue raising capacity per capita
Other Revenue	Per capita total county revenue except property taxes and sales taxes (in thousands of dollars)
Family Income	Median household income (in thousands of dollars)
Total Income	Total income of county residents (in millions of dollars)
Urban	If the county is classified as urban (0 = no, 1 = yes)
Suburban	If the county is classified as suburban (0 = no, 1 = yes)
Rural	If the county is classified as rural (0 = no, 1 = yes)
Tourism Rich	If the county is classified as tourism rich (0 = no, 1 = yes)
Over 65	Percentage of the population over the age of sixty-five
Unemployed	Percentage of the population classified as unemployed
Density	Population density; population per square mile
LOST pc	Per capita local sales tax revenue generated by Article 39
Mills	Millage rate (cents per US$1,000)
Property Value	Per capita total property valuation (in thousands of dollars)
LOST	Local sales tax revenue generated by Article 39

Note. RRC = Revenue raising capacity; LOST = local options sales tax.

Results

Table 4 presents the results for all four steps in this analysis; the dependent variable is LOST revenue per capita for steps 1 through 3 and RRC for step 4. I use a fixed effects model for step 1 and an SDEM for steps 2 through 4. The results of step 1, reported in the first column, reveal a statistically significant difference between the LOST revenue of urban, rural, and suburban counties. Both urban and rural counties are expected to generate more than their suburban counterparts, US$5.23 and US$9.78 per capita, respectively. However, if there is spatial autocorrelation, the residuals will not be independent and identically distributed, and the estimates will be biased (Anselin and Arribas-Bel 2011). After the possibility of spatial heterogeneity and dependence is introduced (column 2), urban counties are estimated to generate roughly US$5.81 more per capita, per annum, than suburban counterparts. However, they also show that rural counties generate US$9.83 more per capita than suburban counties. Given that the average per capita LOST revenue is US$70.71, these effects are substantial. An important change with the new model is that there is less confidence in these relationships. The estimates from the first model were statistically significant at the 1 percent level, whereas they are only significant at the 10 percent level in the second. Also, as expected, the urban spatial lag has a large negative effect. If all of the neighboring counties are urban (i.e., a one-unit increase in the lag), the estimated effect is a US$21.23 reduction in LOST revenue per capita. This supports the expectation that citizens of counties with urban neighbors will take advantage of retail agglomerations and export tax dollars to their urban neighbors. The finding is statistically significant at the 1 percent level.

Table 4.

Fixed Effects Estimation of the Urbanism of Counties on LOST Revenue and RRC.

	Step 1	Step 2	Step 3	Step 4
	LOST per capita	LOST per capita	LOST per capita	RRC
Mills	−0.508	−0.394	−0.374
Mills	(0.284)*	(0.214)*	(0.214)*
Other Revenue	−0.358	−0.458	−0.455	4.165
Other Revenue	(0.554)	(0.708)	(0.707)	(23.704)
Family Income	0.045	−0.055	−0.063	−1.449
Family Income	(0.137)	(0.120)	(0.120)	(4.056)
Total Income	0.010	0.010	0.009	−0.134
Total Income	(0.011)	(0.005)**	(0.005)*	(0.170)
Urban	5.228	5.811	5.924	15.145
Urban	(0.689)***	(3.099)*	(3.094)*	(103.734)
Rural	9.783	9.829	10.136	48.785
Rural	(2.350)***	(5.422)*	(5.415)*	(181.355)
Tourism Rich			13.714	72.584
Tourism Rich			(5.912)**	(197.897)
Income_Spatial Lag		0.359	0.398	47.227
Income_Spatial Lag		(0.244)	(0.244)	(8.081)***
Urban_Spatial Lag		−21.234	−20.995	77.501
Urban_Spatial Lag		(7.951)***	(7.938)***	(266.325)
Unemployed	1.722	1.716	1.725	22.253
Unemployed	(0.659)***	(0.449)***	(0.448)***	(15.031)
Over sixty-five	1.779	1.717	1.703	77.934
Over sixty-five	(0.945)*	(0.518)***	(0.517)***	(17.315)***
Density	−0.025	−0.015	−0.011	0.73
Density	(0.043)	(0.022)	(0.022)	(0.737)
Constant	29.264
Constant	(15.719)*
Observations	700	700	700	700

Note. The standard errors are below the estimated coefficients in parentheses. All regressions include fixed effects for year and county. All counties are classified as one of the following: urban, suburban, rural, or tourism rich. Suburban is the baseline. Step 1 is modeled using fixed effects with a maximum likelihood estimator. Steps 2 to 4 are modeled using a spatial Durbin error panel model (SDEM). RRC = revenue raising capacity.

Asterisks denote significance at the 10 percent (*), 5 percent (**), and 1 percent (***) levels.

Once the suburban and rural counties that are tourism rich are reclassified (step 3), the model estimates indicate that tourism rich counties generate the most LOST revenue per capita, as expected. I report these results in the third column of table 4. If a county is tourism rich, it generates an estimated US$13.71 more in LOST revenue per capita than a suburban county, whereas urban counties generate US$5.92 more and rural counties generate US$10.14 more than average suburban counties. All of the urbanism measures are statistically significant, though only tourism rich is significant at the 5 percent level. Once again, the urban spatial lag is statistically significant, and the estimated effect is a US$21 reduction in LOST revenue per capita. These results support the hypothesis that the biggest winners are tourism rich counties, that the biggest losers are suburban counties and that there is a penalty for being neighbors with an urban county. That rural counties would generate more LOST revenue per capita than their urban counterparts, all else being equal, is unexpected.²⁴ However, since no counties classified as urban have urban neighbors, the expected penalty of having urban neighbors affects rural counties—not urban—and the difference between rural and urban may not be as stark as it appears to be. This also suggests that suburban counties have even less LOST RRC than it appears because, by definition, they have urban neighbors.

The millage rate also has a statistically significant estimated effect. The model estimates that an increase of one mill will bring a LOST revenue reduction of between US$0.51 and US$0.37 per capita. This is in keeping with the literature suggesting that property taxes and LOSTs are substitutes for each other (Ghaus 1995). Nonproperty and sales tax revenues do not affect the LOST revenue. Population density does not have a statistically significant effect on LOST revenue per capita. A one-point increase in the percentage of the population that is over sixty-five also increases LOST revenue per capita by an estimated US$1.72. Notably, the spatial lag of median income is not statistically significant. Contrary to expectations, the median income of county i does not affect its LOST revenue. However, an increase in total resident income of US$1 million is expected to increase LOST revenue per capita by US$0.01.²⁵

Step 4 examines overall RRC, which is measured as the average tax rates multiplied by the tax bases (property taxes and sales taxes) within the county. It is important to acknowledge that property tax revenue represents the largest component of this capacity. The revenue generated by the two is not comparable: an average of US$494 per capita for property taxes compared with US$71 per capita for LOSTs. However, this is in keeping with the literature and an accurate way of assessing the overall equity created by these two taxes.

Counter to expectations, there is no statistically significant difference distinguishing urban, rural, and tourism rich counties from their suburban counterparts. Referring back to table 2, it appears that there are large differences between the per capita RRC. However, diving deeper into the data, this becomes less surprising. The range of per capita RRC is similar in each county categorization. The ranges are as follows: US$360.68 to US$1,175.00 for urban, US$354.15 to US$2,419.52 for suburban, US$250.22 to US$2,147.46 for rural, and US$348.42 to US$3,636.66 for tourism rich. The urbanism designations better explain variability in LOST revenue per capita than in RRC.

The remaining relationships between the independent variables and RRC are mostly as expected. The greater the percentage of the population over sixty-five, the higher the county’s RRC is expected to be. A one-percentage point increase affects the expected RRC by US$77.93 per capita. The wealthier a county’s neighbors, as expressed in the median income spatial lag, the expected RRC is expected to increase by US$47.23 per capita. Interestingly, median income does not have a statistically significant effect on RRC. These results suggest that RRC may not be as inequitable as often assumed, at least with regard to a county’s urban classification. This is in keeping with Zhao and Hou (2008).

Conclusion

Scholars who study LOSTs both assume and find that urban local governments benefit from the use of LOSTs much more than their nonurban counterparts (Rogers 2004; Zhao and Hou 2008; Burge and Rogers 2011; Burge and Piper 2012). For example, Burge and Rogers (2011, 57)²⁶ “find evidence to support the idea that rural communities may be disproportionately constrained when states rely heavily upon LOST revenues for local finance.” Although the urban classification of governments is not the focus of their analysis, studies of LOSTs commonly control for a county being urban and assert or find that urban counties generate less LOST revenue than their rural counterparts. The results here suggest that the real winners are tourism rich counties, which not only generate the most LOST revenue per capita but also have substantial property tax bases. While the results suggest that rural counties do not face the disadvantage often presumed, it finds that suburban counties are likely the biggest tax exporters based on both the urbanism classifications and the urban spatial lag. This is in keeping with the hypotheses and previous literature (Burge and Rogers 2011).

The perception that urban jurisdictions generate an inequitable share of LOST dollars, that suburban communities may be exporting their LOST dollars to these urban communities, and that rural areas do not have the base to generate significant amounts of LOST revenue fuel equity concerns in the literature, county courthouses, and state legislatures. However, more nuanced arguments recognize the additional costs associated with day commuters and tourists in urban areas who use services but do not pay property taxes (Lewis and Barbour 1999; Artz and Stone 2003). Previous research has shown that rural areas, if far enough removed from urban centers, may themselves be regional retail centers (Rogers 2004; Zhao and Hou 2008; Burge and Rogers 2011) and that suburban communities may be the worse off in terms of LOST receipts (Rogers 2004). Suburban jurisdictions may also be extremely affluent areas that can generate a great deal of property tax revenue (Artz and Stone 2003). These reasonable, but somewhat contradictory, views are all important when considering equity. The results presented here suggest that urban counties do not collect more LOST revenue per capita than rural and tourism rich counties and that, in fact, they may actually collect less. The results also suggest that rural areas farther removed from urban counties generate more LOST revenue per capita than those that directly neighbor urban counties.

The practical implications of LOST revenue being unevenly distributed among local governments depend on overall RRC. The results suggest that no statistically significant differences in RRC exist between the different counties based on level of urbanism. Of course, this does not imply that per capita costs for similar services are the same. That would require a more holistic analysis of financial capacity and integration of expenditures into the analysis of equity. This is because the per capita cost of providing services is U-shaped due to the inability of low population jurisdictions with high geographic dispersion to capture economies of scale (Bel and Mur 2009; Burge and Piper 2012).

The results of this study present evidence that state-level LOST policies may inadvertently create a system that puts some communities at a great disadvantage. If a LOST is simply a tool for generating additional revenue, this may be considered acceptable. However, if LOST revenue is being used as a substitute for state funds or raises restrictions on other tax instruments, this may be problematic and create equity concerns.²⁷ This may prompt other states to consider policies like those being considered in N.C. currently, to distribute LOST revenue on a per capita basis.

Footnotes

Acknowledgments

The author would like to thank Daniel Smith, David Ammons, Esteban Afonso, participants at the 2013 annual conference of the Association of Budgeting and Financial Management, the three anonymous reviewers, and James Alm for their helpful feedback and suggestions.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Supplemental Material

The online appendices are available at .

References

Abdel-Rahman

Hesham M.

1990. “Agglomeration Economies, Types, and Sizes of Cities.” Journal of Urban Economics 27:25–45.

ACIR (Advisory Commission on Intergovernmental Relations). 1986. Changing Public Attitudes on Governments and Taxes. Pub. No.S-15. Washington, DC: ACIR.

Afonso

Whitney B.

2013. “Diversification toward Stability? The Effect of Local Sales Taxes on Own Source Revenue.” Journal of Public Budgeting, Accounting and Financial Management 25:649–74.

Afonso

Whitney B.

2014. “Local Sales Taxes as a Means of Increasing Revenues and Reducing Property Tax: An Analysis Using Propensity Score Matching.” Public Budgeting and Finance 34:24–43.

Anselin

Luc

. 1995. “Local Indicators of Spatial Association—LISA.” Geographical Analysis 27:93–115.

Anselin

Luc

Arribas-Bel

Daniel

. 2011. “Spatial Fixed Effects and Spatial Dependence.” RePEc Paper No. 1045, GeoDa Center for Geospatial Analysis and Computation, Tempe, AZ.

Artz

Georgeanne

Stone

Kenneth E.

. 2003. “An Analysis of the Transfer of Funds from Weak Retail Counties to Strong Retail Counties in Iowa via Local Option Sales Taxes.” Paper presented at the Annual Meeting of American Agricultural Economics Association, Montreal, Quebec, Canada.

Baltagi

Badi H.

Yen

Yin-Fang

. 2014. “Hospital Treatment Rates and Spillover Effects: Does Ownership Matter?” Regional Science and Urban Economics 49:193–202.

BEA (Bureau of Economic Analysis). 2015. Income (Place of Residence) and Earnings (Place of Work) by Industry. Washington, DC: U.S. Bureau of Economic Analysis.

10.

Bel

Germà

Mur

Melania

. 2009. “Intermunicipal Cooperation, Privatization and Waste Management Costs: Evidence from Rural Municipalities.” Waste Management 29:2772–78.

11.

Burge

Gregory S.

Piper

Brian

. 2012. “Strategic Fiscal Interdependence: County and Municipal Adoptions of Local Option Sales Taxes.” National Tax Journal 65:387–416.

12.

Burge

Gregory S.

Rogers

Cynthia L.

. 2011. “Local Option Sales Taxes and Consumer Spending Patterns: Fiscal Interdependence under Multi-tiered Local Taxation.” Regional Science and Urban Economics 41:46–58.

13.

Carroll

Deborah A.

2009. “Diversifying Municipal Government Revenue Structures: Fiscal Illusion or Instability?” Public Budgeting and Finance 29:27–48.

14.

Cornia

Gary C.

Grimshaw

Scott

Nelson

Ray

Walters

Lawrence

. 2010. “The Effect of Local Option Sales Taxes on Local Sales.” Public Finance Review 38:659–81.

15.

Craft

Matthew M.

2002. “Lost and Found: The Unequal Distribution of Local Option Sales Tax Revenue among Iowa Schools.” Iowa Law Review 88:199.

16.

Department of Housing and Urban Development. 2010. Income Estimates (North Carolina). Raleigh, N.C. Accessed November 20, 2013, http://portal.hud.gov/hudportal/HUD?src=/states/north_carolina.

17.

Drenkard, Scott. 2013. State and Local Sales Tax Rates in 2013. Washington, DC: Tax Foundation. Accessed March 25, 2014, http://taxfoundation.org/article/state-and-local-sales-tax-rates-2013.

18.

Golgher

Andre

Voss

Paul

. 2014. How to Interpret the Coefficients of Spatial Models: Spillovers, Direct and Indirect Effects. Working Paper.

19.

Ghaus

A. F. Aisha

. 1995. “Optimal Local Sales Tax.” Urban Studies 32:1369–81.

20.

Hausman

Jerry A.

1978. “Specification Tests in Econometrics.” Econometrica 46:1251–71.

21.

Hou

Yilin

Seligman

Jason S.

. 2008. “LOST Stability? Consumption Taxes and the Cyclical Variability of State and Local Revenues.” In Proceedings 100th Annual Meeting of the National Tax Association, Vol. 100, 467–78. Washington, DC: National Tax Association.

22.

Ingram

Deborah D.

Franco

Sheila J.

. 2012. “NCHS Urban-rural Classification Scheme for Counties.” In Vital and Health Statistics, Series 2, No. 154. Washington, DC: U.S. Department of Health and Human Services.

23.

Jung

Changhoon

. 2001. “Does the Local-option Sales Tax Provide Property Tax Relief? The Georgia Case.” Public Budgeting and Finance 21:73–86.

24.

Lawrence

David M.

Millonzi

Kara A.

. 2007. “Revenues. (Article 13).” In County and Municipal Government in North Carolina, edited by Lawrence

David M.

, 1–16. Chapel Hill: UNC School of Government.

25.

Lewis

Paul George

Barbour

Elisa

. 1999. California Cities and the Local Sales Tax. San Francisco: Public Policy Institute of California.

26.

Merryman

Scott

. 2008. USSWM: Stata Module to Provide US State and County Spatial Weight (Contiguity) Matrices. Chestnut Hill, MA: Boston College.

27.

Minkoff

Scott L.

2009. “Minding Your Neighborhood: The Spatial Context of Local Redistribution.” Social Science Quarterly 90:516–37.

28.

Moran

Patrick A. P.

1948. “The Interpretation of Statistical Maps.” Journal of the Royal Statistical Society, Series B (Methodological) 10:243–51.

29.

North Carolina Department of Commerce. 2013a. Travel Economic Impact Model. Raleigh, N.C. Accessed October 3, 2013, www.nccommerce.com/tourism/research/economic-impact/teim.

30.

North Carolina Department of Commerce. 2013b. Labor Force and Employment–Annual. Raleigh, N.C. Accessed October 3, 2013, www.nccommerce.com.

31.

North Carolina Department of Revenue. 2013a. Effective Dates of Local Sales and Use Tax Levies in North Carolina Counties. Raleigh, N.C. Accessed October 3, 2013, www.dornc.com/publications/archive/archive.html.

32.

North Carolina Department of Revenue. 2013b. County revenue data (provided upon request). Raleigh, N.C. Accessed November 12, 2013, http://www.dornc.com/localgovt/index.html.

33.

North Carolina Department of State Treasurer. 2013. Local Government Revenue by Source. Raleigh, N.C. Accessed October 3, 2013, www.nctreasurer.com/Pages/default.aspx.

34.

North Carolina Office of the Governor. 2013. Population and Migration. Raleigh, N.C. Accessed October 11, 2013, www.governor.state.nc.us.

35.

Oates

Wallace

. 1991. “On the Nature and Measurement of Fiscal Illusion: A Survey.” In Studies in Fiscal Federalism, edited by Oates

W. E.

, 431–48. Cheltenham, UK: Edward Elgar.

36.

Rogers

Cynthia L.

2004. “Local Option Sales Tax (LOST) Policy on the Urban Fringe.” Journal of Regional Analysis and Policy 34:25–50.

37.

Rork

Jonathan C.

2000. “Neighboring Influences in State Tax Determination.” In Proceedings: Annual Conference on Taxation and Minutes of the Annual Meeting of the National Tax Association, Vol. 93, 291–94. Washington, DC: National Tax Association.

38.

Schwartz

Jonathan

. 1997. “Prisoners of Proposition 13: Sales Taxes, Property Taxes, and the Fiscalization of Municipal Land Use Decisions.” Southern California Law Review 71:183–217.

39.

Sjoquist

David L.

Walker

Mary Beth

Wallace

Sally

. 2005. “Estimating Differential Responses to Local Fiscal Conditions: A Mixture Model Analysis.” Public Finance Review 33:36–61.

40.

Sjoquist

David L.

Smith

William J.

Walker

Mary Beth

Wallace

Sally

. 2007. “An Analysis of the Time to Adoption of Local Sales Taxes: A Duration Model Approach.” Public Budgeting and Finance 27:20–40.

41.

Tax Policy Center. 2006. Sales Tax Rates 2004. Washington, DC. Accessed November 10, 2013, www.taxpolicycenter.org/TaxFacts/TFDB/TFTemplate.cfm?Docid=411&Topic2id=90.

42.

Tosun

Mehmet Serkan

. 2003. “Cross-border Effects of Alcohol Sales and Intrastate Spatial Tax Incidence.” In Proceedings: Annual Conference on Taxation and Minutes of the Annual Meeting of the National Tax Association, Vol. 96, 410–19. Washington, DC: National Tax Association.

43.

Wachs

Martin

. 2003. “Local-option Transportation Taxes: Devolution as Revolution.” State Tax Notes 28:9–15.

44.

Wang

Wen

Zhao

Zhirong

. 2011. “Fiscal Effects of Local Option Sales Taxes on School Facilities Funding: The Case of North Carolina.” Journal of Public Budgeting, Accounting and Financial Management 23:507–33.

45.

Webb

Myrt

Anderson

Melinda

. 2005. Where We Live: The Impact of How Communities Grow. Washington, DC. Accessed May 20, 2014, www.bookstore.icma.org/freedocs/mlgs/Where%20We%20Live.pdf.

46.

Zhao

Zhirong

. 2005. “Motivations, Obstacles, and Resources: The Adoption of the General-purpose Local Option Sales Tax in Georgia Counties.” Public Finance Review 33:721–46.

47.

Zhao

Zhirong

Hou

Yilin

. 2008. “Local Option Sales Taxes and Fiscal Disparity: The Case of Georgia Counties.” Public Budgeting and Finance 28:39–57.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.21 MB