Market Potential and Fiscal Incentives Influence Firms’ Location Decisions: Evidence From U.S. Counties

Abstract

How important are market potential and fiscal incentives for firms’ location decisions? We estimate the influence of subsidies and tax breaks on the decisions of firms to relocate or to remain in a certain U.S. county using a structural economic geography model developed in Meurers and Moenius (2018). In a panel data set from 1990 to 2016 for almost 3,000 U.S. counties, the authors find a strong and robust impact of economic geography on firms’ location decisions: The closer a county is to market demand and to the supply of inputs, the more firms locate there. As the model predicts, public investment attracts firms while the local tax burden disincentivizes economic activity, although to a lesser extent. Furthermore, in counties that are closer to economic centers, firms respond less to public investment and tax changes than firms in counties far away from centers. These data, therefore, confirm the predictions of the model regarding the potential effectiveness of regional development policies, in particular for investment tax credits, job creation, and training.

Keywords

regional economic policy fiscal incentives market potential

Economic theory suggests that firms choose their location to maximize profits. This requires them to locate close to market demand, to factors of production, or to inputs. The economic geography literature measures proximity to market demand as the distance-discounted sum of (real) economic activity in neighboring regions (see, e.g., Harris, 1954; Krugman, 1991). As transport costs fall, the distance to markets becomes relatively less important, and other factors such as local fiscal incentives increase in importance. In this article, we study this trade-off between access to markets and fiscal incentives, which can take the form of tax exemptions or a high-quality provision of public infrastructure (e.g., road transport or broadband connections).

Previous work has focused on the empirical role of either of these two factors. Our own prior work establishes the importance of localized pecuniary spillovers that may lead to equilibrium levels of investment in public infrastructure near economic centers that are below the social optimum. Henceforth, we will label this phenomenon as underinvestment in the periphery. Our model is static in nature and assumes instantaneous adjustment. As such, it cannot capture the rich, dynamic adjustment processes that are ongoing on a region’s path toward such an equilibrium. However, it delivers some inherently interesting insights, as well as testable hypotheses, that deliver answers to questions of local public policy relevance. First, if underinvestment is an equilibrium outcome in which pecuniary spillovers create attractiveness for firms that compensate for lower public investment in the periphery, underinvestment can likely persist for a long time or repeatedly occur. As such, it should be measurable in the data. Second, underinvestment does not imply no public investment at all. It only indicates lower investment than in the centers in the amount that positive spillovers from neighbors can compensate for lower public investment that allows regions to enjoy lower taxation at the same time. Third, differences in fiscal policies across regions matter for attracting firms.

Therefore, we investigate how market potential, localized incentives, and potential underinvestment in infrastructure jointly determine firm location within the United States over time. The rich U.S. data infrastructure, specifically at the county level with its sizeable public investment and measurable spillover effects, lends itself to this kind of analysis.

We use the general equilibrium framework developed in Meurers and Moenius (2018) to guide our empirical analysis as we identify the relative importance of economic geography and fiscal incentives. We study the effect of local public investment and taxation on the net number of firms in manufacturing. This number includes the founding of new firms net of firm deaths, the firms that persist and stay, and net migration of firms. Our dependent variable is the proportion of total manufacturing firms in a county and we control for the firm size distribution in most of our specifications. Our independent variables capture local fiscal incentives, taxation, and variables that reflect economic geography, such as how much the distance from centers affects relative prices. We control for local factors that are or can be viewed as constant over time with county-specific fixed effects. We control for influences that affect all locations in a similar manner over time with time fixed effects. We expect fiscal incentives to increase the number of firms and taxes to decrease the number of firms. Our theoretical model also predicts that positive spillovers from centers into the surrounding areas leads to an increase in the number of firms in those surrounding areas.

Our panel regressions show that both economic geography and fiscal variables exhibit significant influence on the development of the number of firms at the county level in the expected way. Most notably, our economic geography variables consistently appear large and significant in our results: For firms, it pays to be close to a center and counties can afford to underinvest in public infrastructure there without fending off firms. The detailed available data, moreover, allow us to identify the most effective forms of tax incentives to attract firms: investment tax credits, job creation tax credits, and job training grants. These results have some important consequences for public policy: (1) Both public investment and lower taxes help increase the number of firms; (2) Both types of fiscal incentives are more effective in attracting firms and supporting firm growth in remote places, as there is less leakage through positive spillover effects to competing firms in neighboring counties; and (3) While fiscal incentives seem to be effective to promote firm location and growth, their efficiency is less clear. The magnitude of our estimated effects is rather moderate, such that we do not obtain convincing evidence that such policies might pay off from a taxpayer’s perspective.

Our most pervasive empirical results come from estimating the effect of fiscal policy on the number of firms in centers with high market potential versus the periphery with low market potential. For given market potential differentials between centers and the periphery, fiscal policy is almost meaningless in centers but highly effective to keep or attract manufacturing firms in regions far from centers. As the market potential of a region depends on distance-weighted economic activity in all other regions, the effectiveness of such policies can change over time, namely when transportation costs fall. Therefore, locations that see changes in their market potential need to adjust their fiscal policies accordingly. Our results, however, suggest that economic development support, be it through tax credits or public investment, should be focused on areas distant from the centers, as it has the strongest effect in areas with low market potential.

The article begins with a brief review of the literature and discusses previous empirical findings on the role of market potential and fiscal incentives on firm location. The next section derives the estimation equation from the model in Meurers and Moenius (2018), followed by a description of the data and the calculation of market potential and the adjusted number of firms. We then present the results and conclude with a discussion of policy implications.

Literature Review

Since there is a vast literature on the determinants of firm location, we focus only on the part of the literature most closely related to our work. Brülhart (1998) categorized the factors that affect location choice into two main groups. The first, neoclassical factors such as natural endowments, transportation costs, and technologies are exogenous to the firm at any point in time. The other group comprises endogenous factors, which include spillovers between firms and markets. For the United States, Ellison and Glaeser (1999) provided evidence on the relevance of spillover effects. They attributed at least one fifth, but no more than one half, of industry location choice in the United States to observable cost-determining natural advantages. They conjectured that the remaining—larger—share must be explained by endogenous factors, such as localized intraindustry spillovers, and confirm this in Ellison et al. (2010).

Already Marshall (1890) recognized spillovers resulting from firms at the same location (e.g., through technology sharing or leakages, knowledge transfer, labor pooling, and intermediate input linkages). These are frequently referred to as agglomeration economies. Spillovers, however, can also result from changes in the structure of markets with imperfect competition and increasing returns to scale. Location of firms then heavily depends on the proximity to market demand and to supply factors of production (see Fujita et al., 2001; Krugman, 1991). Identifying the central role of the location of production, its agglomeration in space as evidenced by cities as economic centers as well as the resulting market environment, shapes the foundation of the new economic geography. If production and consumption are concentrated, access to those markets becomes key, with producers and consumers located more closely to an economic center assumed to have less costly access. Economic geography models capture proximity to the relevant market environment with a variable, labeled “structural market access.” This variable closely resembles the concept of market potential by Harris (1954). Typically, indicators of local economic activity, such as local nominal wages and the size of the industrial sector, are positively related to market potential. Firm location is a discrete choice: A firm either locates in a particular location or elsewhere. McFadden (1973) developed a discrete choice modeling strategy for empirical analysis in which individual cases of firm location are observed over a certain time span. Crozet et al. (2004) used this approach to investigate the location choice of foreign investors in France. They found that foreign firms in sectors like computers, car parts, machine tools, and office machinery prefer existing agglomerations of competitors. Over time, however, existing firm clusters lose their relevance and the impact of market potential becomes larger. Interestingly, they did not find any impact of French or European regional policies on the location choice of foreign firms.

Head and Mayer (2004) investigated whether Japanese affiliates in Europe chose locations that offer market access à la Krugman or access to an industry cluster of firms within the same industry. They found that both market potential and industry-specific agglomeration of firms, measured by the share of domestic firms in the sector within a region, helped to explain why firms locate in a specific region.

The evidence therefore suggests that agglomeration positively affects firm location through two different channels: a direct spillover effect between neighboring firms in the same sector and a market-size effect. We add to these two main pillars of this literature in two distinct ways. First, we explicitly capture the channel of the market-size effect; namely, through the effect on relative prices for both consumers and producers who require intermediates for production. Second, we revisit the largely understudied area of fiscal incentives and public policy in the context of location choice.

There is a substantial literature on the economic effects of fiscal policies at the state and local levels. Earlier studies for the United States are surveyed in Bartik (1991). This strand of research empirically assesses how fiscal variables influence local economic outcomes, such as output and employment. Since the importance of market potential has been only recently established by the economic geography literature, these studies use control variables that do not include market potential. In contrast, empirical studies of location choice rarely analyze fiscal incentives. An exception are papers along the lines of Holmes (1998), or Chirinko and Wilson (2008). These studies viewed relatively easy firm relocation between neighboring counties on both sides of a state border as the source that allowed them to identify the effects of differences in fiscal policies on firm location choices. In these studies, policy differences are typically measured at the state level, while firm location choice is recorded at the county level, which gives the policy variables a more exogenous character. However, Peltzman (2016) highlighted that the moderate fiscal policy impact regularly obtained from the difference estimation for border counties might be blurred by demand spillovers resulting from state policies on the other side of the border. Taking these into account, he showed that the true policy impact might be larger than previously recorded. For example, fiscal expansions on both sides of the border diminishes the relative importance of firm-level incentives and thus reduces their effect on firm location choice. Many of the earlier studies on fiscal incentives struggled with problems of multicollinearity of fiscal variables that lead to insignificant results. Specifically, revenues and expenditures are not orthogonal, but depend on each other through budget constraints. Including both types of variables in the same regression generates multicollinearity. As a solution, Helms (1985) suggested to omit expenditure categories from the set of explanatory variables and then interpret the coefficient on the tax variable as the impact of the omitted expenditures on growth as those taxes were raised to finance them. A recent comprehensive survey by Rickman and Wang (2018) noted that despite all advances in econometric techniques (e.g., accounting for multicollinearity, nonlinearity, spatial spillovers, and endogeneity, and despite much better data availability), there is still no clear answer, as the results still vary across studies and strongly depend on individual circumstances in time and space. In particular, Rickman and Wang (2018) argued that there is still no convincing evidence that local tax burden is a major driver of local economic performance in either direction. In contrast, public expenditure particularly expenditure on education and transport seem to have a positive impact on economic growth.

Our article fills the gap between the studies of agglomeration and market potential on firm-specific location decisions and studies that assess the impact of fiscal incentives. Following the model developed in Meurers and Moenius (2018), we investigate how market potential and fiscal incentives jointly determine firms’ location decisions. To do so, instead of individual firm’s location choice, we model the aggregate outcome of individual location choices. This encompasses firm continuation, relocation of firms, and the birth and death of firms.

Theoretical Background

In Meurers and Moenius (2018), a new economic geography model was developed based on the previous works of Fujita et al. (2001), Krugman (1991), and Redding (2016). Instead of land, Meurers and Moenius used immobile labor as a locally fixed factor of production. Immobile labor has insufficient incentives to move (e.g., potential real wage increases from moving cannot compensate for total moving costs, which not only includes physical transportation and cost of living adjustments, but also social adjustment costs such as building new social networks or being separated from family and friends). Furthermore, the authors assumed that both mobile and immobile workers are needed for the production of manufactured goods. Finally, fiscal variables are introduced into the model as local public investment expenditure raises total factor productivity of manufacturing firms. It is financed by a proportional local payroll tax such that the local fiscal budget is always balanced.

The equilibrium in the model is determined by spatial real-wage arbitrage of mobile workers. Equation (1) states the equilibrium condition of the model, requiring that mobile workers’ real wage is equal in all regions. Then, the share of total national mobile workers ( $λ_{i}$ ) locating in each region is determined by an index of economic fundamentals in each region relative to an index of fundamentals aggregated across all regions $R$ . These economic fundamentals are the relative endowment with immobile workers ( $θ_{i}$ ), the relative total factor productivity $f_{i,}$ which depends positively on the local public investment ratio to gross domestic product (GDP; $g_{i} = G_{i} / Y_{i}$ ), the relative tax rate ( $t a x_{i}$ ), and the relative terms of trade ( $P_{X_{i}} / P_{M_{i}}$ ).

λ_{i} = \frac{θ_{i} {(1 - t a x_{i})}^{\frac{1}{1 - α γ}} {[\frac{P_{X_{i}}}{P_{M_{i}}} f_{i} (g_{i})]}^{\frac{γ}{1 - α γ}}}{\sum_{j = 1}^{R} θ_{j} {(1 - t a x_{j})}^{\frac{1}{1 - α γ}} {[\frac{P_{X_{j}}}{P_{M_{j}}} f_{j} (g_{j})]}^{\frac{γ}{1 - α γ}}} .

(1)

The parameter $α$ captures the wage share of mobile workers in the total regional wage bill. $γ$ describes the share of each worker’s income that is spent on tradable manufactures. Both parameters increase the relevance of mobile workers location choices for the equilibrium of the model. If there are differences in economic fundamental (e.g., Region 1 has a higher tax rate than Region 2 [ $t a x_{1} > t a x_{2}$ ]), then the difference in mobile labor allocation between the two regions will be more pronounced with rising $α$ and $γ$ .

The terms of trade capture the proximity to markets. Redding and Venables (2004) showed that they are a linear function of both the distance-weighted demand for final manufacturing goods (market access) and the distance-weighted access to the supply of intermediate inputs (supplier access).

For simplicity, firms are assumed to be symmetric with identical cost functions in every region. This implies that the number of firms ( $N_{i}$ ) is a linear function of manufacturing output ( $X_{i}$ ), fixed cost ( $F$ ), and the elasticity of substitution between manufacturing varieties ( $ε)$ : $N_{i} = \frac{X_{i}}{ε F} .$ From an empirical perspective, we would rather allow for varying firm-specific fixed costs that depend on the existing number of firms. This is, for example, the case when it gets more and more expensive to set up a new firm because of limited space. Such increasing fixed costs would imply that every additional firm has to produce a higher output to be profitable: $d X_{i} = ε \frac{\partial F_{i}}{\partial N_{i}} d N_{i}$ . This leads to firm heterogeneity like in the model of Melitz (2003). The relationship between the number of firms and output in such a model is characterized by a function of the form $X_{i} = F_{0} e^{ϕ N_{i}}$ . The parameter $ϕ$ determines how strongly the firm-specific fixed cost rises with additional firm entry. Based on this reasoning, we can derive the following reduced-form equation for the number of local firms $N_{i}$ :

\log N_{i} = c_{0} + \frac{1}{ϕ} \log X_{i}

(2)

The theoretical model does not include other factors that are relevant for the local cost of production (e.g., the supply of land, labor participation, qualification of labor, or climate conditions and topography). Instead, it reduces all local characteristics to local immobile labor $θ_{i}$ , which is unobservable, in reality. To keep our empirical analysis simple, we will work with county-specific fixed effects, which capture local factors that do not change over time. This includes county-specific deviations from the average impact all other right-hand side variables may have. As a consequence, these fixed effects unfortunately also capture at least part of the impact public investment has on our dependent variable. Public investment should be county specific, as we can expect that the marginal productivity of public investment depends on a large set of local factors, such as the size of the local economy, quality and quantity of existing infrastructure, and local educational attainment. To account for general trends in the number of firms, we scale our dependent variable to each county’s share in the total number of manufacturing firms nationally.

After manipulation of Equations (1) and (2), and assuming a log-linear production function of manufacturing production $X_{i}$ depending on equilibrium mobile labor $λ_{i}$ , exogenous immobile labor $θ_{i}$ , and productivity enhancing public investment $g_{i}$ , we obtain the following reduced-form equation for the relative number of firms in each region:¹

\begin{array}{r} \log \frac{N_{i, t}}{N_{t}} = β_{0, i} + β_{1} g_{i, t} - β_{2} (t a x_{i, t} - t a x_{\emptyset, t}) + β_{3} \log \frac{P_{X_{i, t}}}{P_{M_{i, t}}} \\ - β_{4} \log Y_{t}^{r} + β_{5} \log N_{t} + ξ_{i, t} \end{array}

(3)

Apart from the two fiscal variables and the terms of trade, national GDP ( $Y_{t}$ ) and the total number of firms ( $N_{t}$ ) enter as macro variables on the right-hand side of the equation. This results from Equation (1) as local economic activity is determined by local fundamentals relative to the aggregate economic activity in the country. Hence, all else equal, a rise in aggregate economic activity and/or the total number of firms reduces the relative attractiveness of a region, as this implies the region falls behind and other regions attract mobile labor.

Data

Our sample comprises annual data from 1990 to 2016 ( $T = 27$ ). All variables refer to the county administrative borders characterized at the five-digit level of the Federal Information Processing Standard Publication Code. For each year, we have raw observations for R = 3,088 counties. The data sources are summarized in Table 1.

Table 1.

Variables and Descriptions.

Symbol	Variable	Unit	Source
Y	Personal income	US$ thousands	BEA, local GDP data
G	Public investment, sum over Items F and G (construction and other capital outlays), sum over all units 1-5^a	US$ thousands	U.S. Census, annual survey of state and local government finances
Tax	Tax revenues, sum over Items T (Taxes), and over all units	US$ thousands	U.S. Census, annual survey of state and local government finances
P_X/P_M	Terms of trade	index	Author calculations based on Y
N	No. of establishments, for manufacturing (NAICS 31-33) and SIC 20-39 before 1998	count	U.S. Census, county business patterns

Note. BEA = Bureau of Economic Analysis; GDP = gross domestic product; NAICS = North American Industry Classification System; SIC = Standard Industry Code.

The local government units in the data set with a geographical area encompass the county, cities, townships, special districts, and independent school districts, or educational service agencies.

Number of Firms

Our dependent variable is the number of firms in the manufacturing sector in a specific county, $i$ relative to the total number of manufacturing firms in the United States $N_{i} / N$ . Since the average firm size increases with local production, the estimated impact of policies on firm location might be biased in favor of smaller local economies. To compensate for this potential issue, we adjust the number of firms for firm-size differences as follows: We use the firm size categories by employment in the County Business Patterns to compute a national average distribution of county employment across firm sizes. Then, we allocate the total manufacturing employment in each county to the different firm-size categories according to the average national distribution. We finally add up this allocation of firms across all firm-size categories for each county.

Figure 1 demonstrates the procedure to obtain the adjusted number of firms for Los Angeles County in 2016. A larger share of Los Angeles County’s 338,448 manufacturing workers work in smaller firms as compared with the national average. Thus, after adjusting the distribution of employment across firm sizes in Los Angeles County to match the national distribution (Figure 1) and reallocating Los Angeles County total employment to the firm size categories, Los Angeles County has fewer small firms than before adjustment. This leads to a reduction from 12,105 firms before adjustment to 8,569 firms after adjustment. Unless noted otherwise, we use the adjusted values throughout for our analysis.

The County Business Patterns industry classification switched from Standard Industry Code (SIC) to the North American Industry Classification System (NAICS) in 1998. This change in classification leads to small, but measurable, structural shifts in the number of firms. In our empirical analysis, we therefore preemptively account for these structural shifts by including a dummy variable $S I C$ that takes the value 1 before 1998 and 0 thereafter.

Public Investment and Taxes

Local public investment and tax revenues are taken from the U.S. Census of Governments (see Randall et al., 2018, for a detailed description). Our fiscal variables of interest are aggregate taxation and investment expenditures (construction and other capital outlays) at the county level. We aggregate the data at the county level across all five types of county governments: the county itself, the embedded cities, townships, special districts, and the independent school districts. Aggregate tax revenues at the county level contribute roughly 40% of total county revenues. Expenditures on construction and other capital outlays constitute 12% of total county expenditure. The significant residual in the budget allows us to interpret the coefficient on the tax variable as the impact of raising taxes for purposes other than investment. We can also think of the coefficient on public investment to be partially independent from local taxes, as they are at least partially financed through earmarked transfers from the state and the federal levels. The correlation coefficient between our tax and investment variable amounts to 0.52 for our whole sample of 80,728 observations. Thus, we are confident to avoid many of the typical collinearity problems.

County administrations spent about $200 billion in 2016 on construction and other capital outlays, which amounts to about 1.1% of national GDP. The counties’ public investment contributes to approximately 40% of total national public investment. Roughly 80% of counties’ investment expenditures is on construction. The largest investment items are schools (28%), highways (9%), sewerage (8%), water (7%), and public transit (6%). Independent school districts and other special districts perform 45% of all county-level investment, followed by cities (39%) and the county administration (16%).

County tax revenues in 2016 amounted to $610 billion or around 3.3% of national GDP. County tax revenues account for 19% of total U.S. tax revenue. By far the largest portion of counties’ taxes is raised through property taxes (72%). Again, the special and school districts receive the largest share of the tax revenues (42%), followed by cities (33%) and the county administration (25%).

To construct expenditure and tax measures relative to local economic activity, we calculate ratios to personal income at the county level ( $g = G / Y$ and $t a x = T a x / Y$ ), as in Yu and Rickman (2013).

Upjohn Panel Database on Incentives and Taxes

Understanding the joint influence of fiscal policies and economic geography on the number of firms helps policy makers and economic development officers understand their county’s potential. However, from a policy maker’s point of view, it is of particular interest which kind of tax incentives are most promising to promote regional economic development. We therefore test the impact of state level incentives and taxes on firm location. We employ the Panel Database on Incentives and Taxes (PDIT) by the W.E. Upjohn Institute for Employment Research.² The database provides information on how much taxes a business would pay and incentives it would receive if it opened a new facility in a particular year. It includes 33 states covering over 90% of U.S. output for each year from 1990 to 2015. It details this information by five different types: (1) property tax abatements, (2) customized job-training grants, (3) job-creation tax credits, (4) investment tax credits, and (5) research and development (R&D) tax credits. We include these types of incentives separately in our analysis. We use the tax incentives relative to sectoral value added and aggregate overall manufacturing industries.

Since the PDIT data are at the state level, adding the incentive variables into Equation (3) with identical values for all counties in a particular state is of limited value. In our estimations, we therefore apply a variation of the “border approach” sketched in our literature review, and test whether differences in state incentives affect the relative growth of the number of firms in counties in distinct states. To avoid a complex and untraceable interaction between market potential at the county level with policies at the state level, we only match counties in different states that have similar market potential.

Terms of Trade

The terms of trade are computed by iteration. From the price equations in Meurers and Moenius (2018), it follows that $P_{X} / P_{M}$ is a linear function of the ratio of market access ( $M A$ ) to supplier access ( $S A$ ; see also Redding & Venables, 2004):

T o T_{i} = \frac{P_{X_{i}}}{P_{M_{i}}} = φ \frac{{[\sum_{j = 1}^{R} Y_{j} \cdot {P_{M_{j}}}^{ϵ - 1} \cdot {T_{i, j}}^{1 - ϵ}]}^{\frac{1}{ϵ}}}{{[\sum_{j = 1}^{R} Y_{j} \cdot {P_{X_{j}}}^{- ϵ} \cdot {T_{i, j}}^{1 - ϵ}]}^{\frac{1}{1 - ϵ}}} = φ \frac{M A}{S A}

(4)

The factor $φ$ is a combination of the parameters in the cost function of the final goods producers and unknown, a priori. In the model from Meurers and Moenius (2018), each year’s price indices $P_{X_{i}}$ and $P_{M_{i}}$ —and therefore the terms of trade—only depend on variables of the same year. This allows us to calculate them for each year separately. We start with the first year in our sample, 1990, for which we initiate the iteration by setting $P_{M_{i}} = P_{X_{i}} = 1$ . Iceberg transportation costs between regions $i$ and $j$ are calculated as $T_{i, j} = 1 + {d_{i, j}}^{0.35}$ , where $d_{i, j}$ is the Euclidean greater circle distance between the centroids of the counties’ administrative borders. We follow Hummels (1999) and Anderson and van Wincoop (2004) and set the elasticity of trade costs with respect to distance to 0.35. We compute the internal distance within a county following Head and Mayer (2000) as $d_{i, i} = 2 / 3 \sqrt{{Area}_{i} / π}$ . We set the elasticity of substitution between varieties of manufacturing goods to $ϵ = 3.5$ , which is consistent with the trade cost elasticities of Simonovska and Waugh (2014). Based on these assumptions, we compute an initial value of $\frac{M A}{S A}$ . We obtain $\hat{φ}$ from a linear regression (without constant) of $\frac{P_{X}}{P_{M}}$ onto $\frac{M A}{S A}$ and compute the new value for the terms of trade as $\hat{φ} \frac{M A}{S A}$ , which we then can also transform into new estimates for the price indices $P_{X_{i}}$ and $P_{M_{i}}$ . We use these new values to recalculate $\frac{M A}{S A}$ and repeat the procedure until we achieve convergence (i.e., that $\sqrt{\frac{1}{R} \sum_{i = 1}^{R} {(T o T_{i_{n e w}} - T o T_{i_{o l d}})}^{2}} < 0.0015)$ . We repeat the same process for all years.

Note that our terms of trade variable improves on commonly used measures of market potential in several ways: Instead of any ad hoc specification, it is directly derived from the basic specifications of production and consumption in our model. Ad hoc measures of market potential, such as population and population density as well as distance from large cities, can be seen to only capture part of the market potential, as population and population density alone do not capture economic power of a location and distance to just one city ignores the economic potential of all other locations within reach.

The terms of trade calculated in this way for all counties for the year 2016 are displayed in Figure 2. It confirms the well-known spatial distribution of economic activity in the United States with the traditional industrial heartland and large cities on the east, as well as the rising economic centers³ on the West Coast (e.g., Los Angeles County and Silicon Valley).

As terms of trade greater than one (ToT_i > 1) imply that prices of goods exported from a region are higher than those imported into a region, this is positive for local immobile factors that benefit from higher factor income and purchasing power. As mobile labor locates based on real wages, which are the same across locations in equilibrium, there are no direct benefits for mobile labor. As the figure indicates, New York has the highest terms of trade in the country and counties like Glacier County in Montana have the lowest due to its remote location from any other economic activity.

The model in Meurers and Moenius (2018) suggests that firms in counties with lower terms of trade/lower market potential are more likely to respond to fiscal incentives. In particular, the model considers that every fiscal incentive has a price tag in the form of foregone alternative public expenditure or higher taxes. Furthermore, economic activity in centers, including the economic activity fostered by local fiscal policies there, creates positive spillovers for nearby locations through the access to the large market. Therefore, the rational location choice for newly created firms and those on the move is to locate in counties close to central marketplaces where they can enjoy the spillovers from the center while avoiding the higher equilibrium tax rate in the center. As a result, fiscal incentives are less effective in and near economic centers as the spillovers dominate.

In contrast, firms in remote places do not benefit as much—if at all—from positive spillovers originating in other counties. Therefore, fiscal incentives in remote places appear highly effective as they predominantly—if not exclusively—benefit local businesses. To account for this in our empirical analysis, we estimate the impact of fiscal incentives on firm growth for counties with high, intermediate, and low market potential separately.

Descriptive Statistics

Total variance in our dependent and independent variables can be decomposed into the variance across counties and the variance over time. Table 2 first displays summary statistics for the averages across all counties over time (27 observations), then for the averages over time for all counties (about 3,000 observations), and finally across the individual observations both across counties and over time. The standard deviations in Table 2 reveal that for all variables the variation across counties dominates the variation over time. The averages across all counties of local public investment and the tax burden (as a ratio to personal income) both fluctuate over the 27 years without a specific time trend. Both drop to minimum values in the years 1993 to 1996 during a period of economic recovery (see Figure 3).

Table 2.

Summary Statistics Across Time and Across Counties.

Variable	Symbol	Observation	Mean	Median	SD	Min	Max
For averages across all counties
Public investment	$g_{. t}$	27	.0113	.0110	.0028	0.0064	0.0164
Tax burden	$t a x_{. t}$	27	.0294	.0295	.0068	0.0173	0.0411
Terms of trade	$T o T_{. t}$	27	.9068	.9068	.0005	0.9059	0.9076
Share of local firms	$N_{. t} / N_{t}$	27	.00033	.00033	.0000	0.00033	0.00033
For averages across all periods
Public investment	$g_{i .}$	2,994	.0114	.0095	.0107	0.0005	0.4016
Tax burden	$t a x_{i .}$	2,994	.0294	.0257	.0270	0.0024	1.056
Terms of trade	$T o T_{i .}$	3,088	.9068	.9104	.1215	0.3829	1.530
Share of local firms	$N_{i} / N_{i .}$	3,081	.00033	.0001	.0011	0	0.0368
For all individual observations
Public investment	$g_{i, t}$	80,728	.0113	.0081	.0162	0	0.7557
Tax burden	$t a x_{i, t}$	80,728	.0294	.0259	.0326	0	1.924
Terms of trade	$T o T_{i, t}$	83,376	.9068	.9108	.1216	0.3808	1.562
Share of local firms	$N_{i, t} / N_{t}$	81,579	.00033	.0001	.0011	0	0.0471

During that period, apparently the recovery of local government revenues and expenditure lagged relative to other local economic aggregates. Counties with a ratio of investment to personal income close to the sample mean over the 27-year average are, for instance, Scott (Missouri), Lawrence (Ohio), and DeKalb (Georgia). Counties with particularly low investment ratios (1% or even lower) are Buffalo (South Dakota) and Lancaster and Charlotte (both Virginia). Counties with high ratios (9% and above) are Big Stone (Minnesota), Platte (Nebraska), and Aleutians East (Alaska). Counties with a tax burden (local revenues as a ratio to personal income) close to the sample average are Effingham (Georgia), St. Clair (Michigan), Deuel (South Dakota), and Wilson (North Carolina). Counties with a tax burden at the bottom of the sample (0.4% and lower) are, for example, Hale (Alabama), Chattahoochee (Georgia), Oglala Lakota (South Dakota), and Yukon-Koyukuk (Alaska). Counties with a tax burden at the top of the sample (22% and more) are Kennedy and King (both Texas), Eureka (Nevada), and North Slope (Alaska).⁴

Results

Table 3 presents the results of our panel estimations of Equation (3).⁵ We use robust standard errors throughout. In column 1, we present our baseline regression for the raw number of firms, including county-fixed effects only. Market potential is highly significant and affects the number of local firms positively as expected. Fiscal policies carry the expected signs but are not statistically significant.

Table 3.

Panel Regression.

Dependent variable: $\log N_{i, t} / N_{t}$
	I (unadjusted)	II (all counties)	III (center)	IV (periphery)	V (rural)
$\log T o T_{i, t}$ (terms of trade)	2.240*** (0.344)	4.227*** (0.542)	2.955* (1.748)	4.976*** (0.822)	2.082 (2.544)
$t a x_{i, t} - t a x_{\emptyset, t}$ (relative tax burden)	−0.025 (0. 162)	−0.478* (0.267)	−0.146 (0.630)	−0.355 (0.363)	−0.854* (0.462)
$g_{i, t}$ (public investment)	0.187 (0.153)	0.915*** (0.269)	0. 384 (0 .758)	0. 732** (0.352)	1.215*** (0.451)
$\log N_{t}$ (total firms)	−0.037 (0. 039)
$\log Y_{t}$ (national income)	0.299*** (0. 034)
SIC-Dummy	0.118*** (0. 006)	−0.093*** (0.011)	− 0.095*** (0. 019)	−0.105*** (0.014)	0.012 (0.035)
Constant	−0.727 (3.109)	−12.01*** (1.292)	−11.98*** (4.087)	− 8.797*** (1.742)	−35.94*** (7.081)
Time trend	−0.007*** (0. 002)	0.002** (0.001)	0.002 (0.002)	0.000 (0.001)	0.013*** (0.001)
R ²	.097	.179	.152	.084	.106
County fixed effects	yes	yes	yes	yes	yes
Time fixed effects	no	yes	yes	yes	yes
Observation	66,127		15,027	40,938	10,162
Counties	2,870		594	1,741	535

Note. Robust standard errors in parenthesis.

,**, and *** denotes significance at the 1%, 5%, and 10% levels, respectively.

As suggested above, the impact of policies might be blurred if the number of firms changes due to composition effects without altering county employment. For example, Henly and Sanchez (2009) showed that there is a long-run trend starting in the early 1970s that workers become more evenly distributed among establishment types and less concentrated in large firms. Thus, there could be shifts in the local share of firms that are independent of policies. We therefore think of the unadjusted number of manufacturing establishments as measures of innovation and local product and process differentiation, which are important aspects of economic development but do not cover the whole range of it. Henceforth, we therefore switch to the size-adjusted number of establishments.

After switching to the size-adjusted form of our dependent variable, all variables of interest in columns 2 to 5 are of the expected sign. As we will see, parameter values and statistical significance across columns vary in an economically meaningful way. In column 2, we display the results for county-fixed and time-fixed effects. Since model selection tests (Hausman-test, likelihood-ratio for joint significance of effects) favor a model with both county- and time-fixed effects, we base our explanations on the results of this column. We can interpret the coefficients in column 2 as follows: A 1% increase in relative market potential as measured by the terms of trade leads to a more than 4 percentage point increase in a county’s share in the total number of manufacturing firms. Thus, holding the total number of firms constant, this would also imply a rise in the absolute number of firms in a county by 4%. A 1% increase in market potential is equivalent to moving from a county close to the sample mean, such as Oklahoma County to one with a 1% higher market potential, namely, Tulsa County (also in Oklahoma). For Oklahoma County, which had, on average, 730 firms, such a market potential increase would lead to an increase of about 31 manufacturing firms. An increase of market potential by 1 standard deviation (0.12) relative to the sample mean (0.91)—comparable to a move from Oklahoma County to Winnebago County in Wisconsin—would imply a rise in the number of firms by 57%.

A 1 percentage point higher local tax burden, all else equal, leads to a decrease in a county’s share of the total number of manufacturing firms by 0.5 percentage points. For a county at the sample mean in terms of the number of manufacturing firms (130), a 1 percentage point increase in local taxes implies a decrease in the number of local manufacturing firms by roughly one firm—again holding the national number of firms constant. For the Oklahoma County example (with, on average, 730 manufacturing firms), it would mean a loss of four firms. A 1 standard deviation increase, which is equivalent to a local tax hike by 2.7 percentage points, would thus imply a loss of approximately 1.3% of local firms, holding the total number of firms constant. Such a hike or even stronger tax hikes occurred over the 27-year time span in only 103 counties (tax declines of equivalent size occurred in 134 counties).

A 1 percentage point increase in the ratio of public investment to local personal income leads to an increase in a county’s share of the total number of manufacturing firms by 0.9 percentage points. A 1 standard deviation increase in the ratio of public investment (by 1.07 percentage points) leads to a roughly equivalent increase in the number of local firms. A county at the sample mean has about 130 manufacturing firms. For such a county, a 1% increase in the public investment to local personal income ratio implies an increase in the number of manufacturing companies by roughly one firm (an increase by six firms in the Oklahoma case), which is similar to the impact of a 1 percentage point tax hike. Such an increase or more occurred in 348 counties; an equivalent decline in this ratio occurred in 338 counties during the 27 years of our sample.

We repeat our exercise for subsets of counties to study the relative importance of fiscal variables at different levels of market potential. Inspection of the spatial distribution in Figure 2 suggests that while low market potential counties (dark blue) extend over roughly half of the U.S. territory, they constitute rather a minority in the total number of counties. The same applies to economic centers (dark red). To separate economic agglomerations from remote and intermediate distances to central markets, we therefore split the sample into a lower 20%, an intermediate 60%, and an upper 20% quantile with respect to the market potential variable.⁶ We present the results in columns 3 through 5: Column 3 presents the results for the top quantile of counties with the highest market potential (values for terms of trade equal to one or above), column 4 the results for the second quantile (values for terms of trade larger or equal to 0.8 but smaller than 1), and column 5 for the quantile of counties with the lowest market potential (values of the terms of trade smaller than 0.8).

In counties with high market potential (3), coefficients on fiscal policies are smaller in absolute value than in the overall sample and are statistically insignificant, but market potential still has a substantial and statistically significant effect. The opposite is true in counties with small market potential (5). In these counties, both taxes and investment have a larger effect than in the overall sample; market potential has an even lower effect than in the high market potential subsample. Interestingly, counties with intermediate market potential (column 4) show moderate, statistically significant, effectiveness of fiscal policy. However, they exhibit more than twice the effectiveness of market potential on the number of firms than in the low market potential areas. Firms in counties with intermediate market potential appear similarly or slightly less sensitive to the overall tax burden as compared with the total sample, but the estimates are statistically insignificant.

In the next step, we study the effect of specific tax incentives on firm location using the matching procedure described earlier. Recall that we would like to compare similar counties in terms of market potential that are exposed with different policies, which requires them to reside in different states. To select those, we first compute the average market potential of our 2,343 counties from 1990 to 2015 for which we have data on fiscal incentives. On average, we have 70 counties per state, which would allow us to generate 2.6 million matches of counties in distinct states in our sample. To ensure that the selected counties are similar in terms of market potential, we keep only 0.03% of the pairs, namely those pairs with the closest market potential. To achieve this selection, we set an upper tolerance limit for between-county deviations of market potential of 5/10,000 of the standard deviation in average market potential (which is 0.1215). This yields an acceptable number of 689 matching pairs of counties. Table 4 shows some examples of matched counties with high and low terms of trade.

Table 4.

Examples of Matched Counties.

High market potential					Low market potential
Terms of trade	County 1	Adjusted number of firms	County 2	Adjusted number of firms	Terms of trade	County 1	Adjusted number of firms	County 2	Adjusted number of firms
1.090	Hamilton, Ohio	1,613	Oakland, Michigan	1,769	0.797	Mohave, Arizona	78	Carson City, Nevada	80
1.064	Columbia, New York	45	Fulton, Georgia	793	0.789	Eagle, Colorado	8	Sierra, California	1
1.049	Hillsdale, Michigan	118	Milwaukee, Wisconsin	1,670	0.785	Garden, Nebraska	2	Midland, Texas	57
1.026	Clark, Kentucky	78	Forest, Pennsylvania	4	0.769	La Plata, Colorado	18	Torrance, New Mexico	1
1.024	Louisa, Virginia	23	Orange, California	4,497	0.747	Sierra, New Mexico	1	Greenlee, Arizona	1

Note. Terms of trade and adjusted number of firms are averages over the period 1990 to 2016; adjusted noumber of firms are rounded up to the next whole number.

For these pairs, we then compute the differences in all our variables, as well as the fiscal incentives from the PDIT for each year in our sample.

To account for possibly countervailing effects between the overall tax burden and tax incentives, we compute measures for tax incentives relative to the previously employed overall tax burden relative to personal income. For example, for investment tax credits ( $I n v . T C .$ ), we compute our measure for the differences between matched counties $i$ and $j$ as:

Δ {(I n v . T C . - t a x)}_{i j} = (I n v . T C ._{i} - t a x_{i}) - (I n v . T C ._{j} - t a x_{j}) .

(5)

The calculation shows that our measure identifies the differences in specific tax incentives relative to the compared counties’ overall tax burdens. Employing these measures as explanatory variables in our regression then shows how those differences in fiscal variables influence the rate of agglomeration of manufacturing firms across counties. We repeat the fixed-effects regressions from Table 3, in which we replace our fiscal variables with the newly constructed ones on the right-hand side. We also replace county dummies with matching county-pair dummies.

The results with respect to the PDIT fiscal incentives in Table 5 show a differentiated picture with respect to the likely effectiveness of these incentives. In the first column, we repeat the equivalent exercise to column 2 in Table 3. First, note that the fiscal variables have the expected sign, but only the difference in tax rates is marginally significant. In columns 2 to 7, we replace the overall tax burden with differences in specific tax incentives, net of the overall tax burden. Again, note that all incentive variables have the expected sign, but not all are statistically significant: While overall incentives ( $T o t . I n c .$ ), R&D tax credits ( $R & D . T C .$ ), and property tax abatements ( $P r p . T A .$ ) are not significant for differences in firm agglomeration across states, investment tax credits ( $I n v . T C .$ ), job creation tax credits ( $J o b C . T C .$ ) and job training grants ( $J o b T . G r .$ ) are significant at least at the 10% level. The coefficients on these later three fiscal incentive variables (see columns 3, 6, and 7 of Table 5) are all close to 1. This means that across otherwise comparable counties, a 1 percentage point higher tax incentive is associated with a 1 percentage point higher firm agglomeration.

Table 5.

Panel Regression for Matched Counties ij.

Dependent variable: ∆log N_ij,t
	I	II	III	IV	V	VI	VII
$Δ t a x_{i j, t}$ ∆tax _ij,t (Total tax difference)	−0.871* (0.500)
$Δ g_{i j, t}$ ∆g _ij,t (Difference in public investment)	0.666 (0.473)	0.577 (0.463)	0.723 (0.471)	0.626 (0.470)	0.540 (0.467)	0.700 (0.475)	0.670 (0.473)
$c o n s t a n t$	−0.014*** (0.000)	−0.015*** (0.000)	−0.015*** (0.000)	−0.015*** (0.000)	−0.015*** (0.000)	−0.015*** (0.000)	−0.014*** (0.000)
$Δ {(T o t . I n c . - t a x)}_{i j, t}$ (Difference in total incentives		0.544 (0.439)
$Δ {(I n v . T C . - t a x)}_{i j, t}$ (Difference in investment tax credit)			1.061** (0.510)
$Δ {(R & D . T C . - t a x)}_{i j}$ (Difference in R&D tax credit)				0.720 (0.494)
$Δ {(P r p . T A . - t a x)}_{i j}$ (Difference in property tax abatements)					0.407 (0.487)
$Δ {(J o b C . T C . - t a x)}_{i j}$ (Difference in job creation tax credit)						0.901* (0.483)
$Δ {(J o b T . G r . - t a x)}_{i j}$ (Difference in job training grants)							0.889* (0.503)
R ²	.011	.012	.023	.004	.002	.015	.013
Observations	14,217
Counties	689

Note. Robust standard errors in parenthesis. A Hausman test rejects a random effects model at the 1% level.

,**, and *** denotes significance at the 1%, 5%, and 10% levels, respectively.

Four aspects contribute to the noteworthiness of these results. First, across state differences in fiscal incentives are arguably small as compared with other aspects or factors not explicitly included in our estimation, such as quality of life, access to research facilities, or varying transportation costs that depend on the availability of transportation means such as waterways versus trains. Second, certain incentives, specifically workforce-related ones, may have immediate direct effects, while others, such as R&D incentives, may only lead to firm growth in the longer run, which is not captured by our specification. Third, as Figure 3 reveals, fiscal incentives are quite volatile over time; thus, firms may not expect incentives to persist, which may lead to responses considering only the average differences across locations, not to variation over time. Finally, as highlighted by Peltzman (2016), if we match counties that share a common state border (which only occurs unintentionally in our approach), we cannot rule out spillovers that reduce the measured impact despite sharp policy differences. That our specification still finds significant results shows that differences in fiscal incentives are effective in attracting and growing firms.

Discussion and Policy Implications

In contrast to many previous empirical studies that are inconclusive concerning the impact of fiscal policy, we find local investment to be effective in terms of keeping and attracting firms. Our study differs from previous work in that we consider county aggregates of fiscal variables instead of specific tax and expenditure instruments. The effectiveness of those instruments may depend on local circumstances, which are hard to tease out. Considering county aggregates instead provides us with a large data set that allows controlling for general county- and time-specific effects. We consider this as key to obtaining our results.

The results bear substantial significance for policy makers. In a nutshell, they indicate that low market potential counties far away from economic centers benefit the most from local investment to attract manufacturing firms. They also indicate that the same policies are likely less effective in central places—they could be even entirely ineffective there. Moreover, differences in fiscal incentives across states matter for firm agglomeration: Specifically, investment tax credits, job creation tax credits, and job training grants have the potential to attract, keep, and grow manufacturing firms locally.

There are also some important caveats. This study only focused on the manufacturing sector as it is most closely related to the model in Meurers and Moenius (2018). Strictly speaking, all results we obtain can therefore only be taken as guidance for the manufacturing sector. The model also makes predictions about the service sector and could be also applied to other sectors if interpreted loosely. This would be specifically important to provide general guidance for policy as manufacturing continuously decreases in importance, specifically in terms of employment, in the United States. Another issue is that we only measure the direct effect of fiscal policy on manufacturing firms. Due to the large geographic spillover effects present in the data, part of the effect of fiscal policy in the center actually benefits the periphery close to the centers. As these effects are not directly identified in our theoretical model, we also excluded them in the empirical analysis to stay true to the model. We intend to address some of these issues in future work.

Supplemental Material

Mathematical_Appendix – Supplemental material for Market Potential and Fiscal Incentives Influence Firms’ Location Decisions: Evidence From U.S. Counties

Supplemental material, Mathematical_Appendix for Market Potential and Fiscal Incentives Influence Firms’ Location Decisions: Evidence From U.S. Counties by Martin Meurers and Johannes Moenius in Economic Development Quarterly

Footnotes

Acknowledgements

We would like to thank our editor, Dr. Timothy Bartik, for comments that substantially helped clarify our results and thus made them more accessible for readers outside of geographical economics. We would also like to thank two anonymous referees for many valuable suggestions that helped us improve our article. All remaining errors are ours.

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.

ORCID iD

Martin Meurers

Supplemental Material

Supplemental material for this article is available online.

Notes

Author Biographies

Martin Meurers is head of the Division for Fiscal and Business Cycle Policies at the Federal Ministry for Economic Affairs and Energy in Berlin. His interest in spatial analysis of fiscal policy stems from his work on the effectiveness of policies to improve public infrastructure.

Johannes Moenius is the director of the Institute for Spatial Economic Analysis at the University of Redlands. He holds a PhD in economics from the University of California, San Diego. In his research, he is interested in how domestic institutions affect international trade as well as how geography shapes economic decision.

References

Anderson

J. E.

van Wincoop

(2004). Trade costs. Journal of Economic Literature, 42(3), 691-751. https://doi.org/10.3386/w10480

Bartik

T. J.

(1991). Who benefits from State and Local economic development policies? Upjohn Press.

Bartik

T. J.

(2017). A new panel database on business incentives for economic development offered by State and Local governments in the United States. Upjohn Press.

Brülhart

(1998). Economic geography, industry location and trade: The evidence. World Economy, 21(6), 775-801. https://doi.org/10.1111/1467-9701.00163

Chirinko

R. S.

Wilson

D. J.

(2008). State investment tax incentives: A zero-sum game? Journal of Public Economics, 92(12), 2362-2384. https://doi.org/10.1016/j.jpubeco.2008.07.005

Crozet

Mayer

Mucchielli

J.-L.

(2004). How do firms agglomerate? A study of FDI in France. Regional Science and Urban Economics, 34(1), 27-54. https://doi.org/10.1016/S0166-0462(03)00010-3

Ellison

Glaeser

E. L.

(1999). The geographic concentration of industry: Does natural advantage explain agglomeration? American Economic Review, 89(2), 311-316. https://doi.org/10.1257/aer.89.2.311

Ellison

Glaeser

E. L.

Kerr

W. R.

(2010). What causes industry agglomeration? Evidence from coagglomeration patterns. American Economic Review, 100(3), 1195-1213. https://doi.org/10.1257/aer.100.3.1195

Fujita

Krugman

Venables

A. J.

(2001). The spatial economy: Cities, regions, and international trade (Vol. 1). MIT Press.

10.

Harris

C. D.

(1954). The market as a factor in the localization of industry in the United States. Annals of the Association of American Geographers, 44(4), 315-348. https://doi.org/10.2307/2561395

11.

Head

Mayer

(2000). Non-Europe: The magnitude and causes of market fragmentation in the EU. Review of World Economics, 136(2), 284-314. https://doi.org/10.1007/BF02707689

12.

Head

Mayer

(2004). Market potential and the location of Japanese investment in the European Union. Review of Economics and Statistics, 86(4), 959-972. https://doi.org/10.1162/0034653043125257

13.

Helms

L. J.

(1985). The effect of state and local taxes on economic growth: A time series-cross section approach. Review of Economics and Statistics, 67(4), 574-582. https://doi.org/10.2307/1924801

14.

Henly

S. E.

Sanchez

J. M.

(2009). The U.S. establishment size distribution: Secular changes and sectoral decomposition. Economic Quarterly, 95(4), 419-454. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2188482

15.

Holmes

T. J.

(1998). The effect of state policies on the location of manufacturing: Evidence from state borders. Journal of Political Economy, 106(4), 667-705. https://doi.org/10.1086/250026

16.

Hummels

(1999). Towards a geography of trade costs. https://www.krannert.purdue.edu/faculty/hummelsd/research/toward/tgtc.pdf

17.

Krugman

(1991). Increasing returns and economic geography. Journal of Political Economy, 99(3), 483-499. https://doi.org/10.1086/261763

18.

Marshall

(1890). The principles of economics. McMaster University Archive for the History of Economic Thought.

19.

McFadden

(1973). Conditional logit analysis of qualitative choice behaviour. In Zarembka

(Ed.), Frontiers in econometrics (pp. 105-142). Academic Press.

20.

Melitz

M. J.

(2003). The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica, 71(6), 1695-1725. https://doi.org/10.3386/w8881

21.

Meurers

Moenius

. (2018). Optimal public investment in economic centers and the periphery. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3198783

22.

Peltzman

(2016). State and local fiscal policy and growth at the border. Journal of Urban Economics, 95(September), 1-15. https://doi.org/10.1016/j.jue.2016.06.003

23.

Randall

Gordon

Greene

Huffer

(2018). Follow the money: How to track federal funding to local governments. https://www.urban.org/sites/default/files/publication/96761/2018.02.26_follow_the_money_v4_-_printpdf.pdf

24.

Redding

(2016). Goods trade, factor mobility and welfare. Journal of International Economics, 101(July), 148-167. https://doi.org/10.1016/j.jinteco.2016.04.003

25.

Redding

Venables

A. J.

(2004). Economic geography and international inequality. Journal of International Economics, 62(1), 53-82. https://doi.org/10.1016/j.jinteco.2003.07.001

26.

Rickman

D. S.

Wang

(2018). U.S. State and local fiscal policy and economic activity: Do we know more now? https://mpra.ub.uni-muenchen.de/88422/1/MPRA_paper_88422.pdf

27.

Simonovska

Waugh

M. E.

(2014). The elasticity of trade: Estimates and evidence. Journal of International Economics, 92(1), 34-50. https://doi.org/10.1016/j.jinteco.2013.10.001

28.

Rickman

D. S.

(2013). US state and local fiscal policies and non-metropolitan area economic performance: A spatial equilibrium analysis. Papers in Regional Science, 92(3), 579-597. https://doi.org/10.1111/j.1435-5957.2012.00423.x

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

0.11 MB