Local Fiscal Policies and Small Businesses in Florida:

A Spatial Panel Analysis

Hai (David) Guo

Associate Professor

Department of Public Administration

Steven J. Green School of International and Public Affairs

Florida International University

and

Shaoming Cheng

Associate Professor

Department of Public Administration

Steven J. Green School of International and Public Affairs

Florida International University

Abstract: This paper is intended to examine the effects of local option taxes and

local fiscal expenditures on small businesses in Florida. As a case study focusing

only on one state and on the manufacturing sector, this paper represents a pilot

study for shedding light on the linkage between small business development and

local fiscal decisions, which seem to have no obvious or direct impact on business

assistance. Spatial panel regression models are calibrated with county-level tax,

expenditure, and social and economic information for the period of 2008-2012. It

is suggested that local tax and expenditure structure and decisions affect the stock

of small manufacturing business establishments not only in their “home” counties

but also in their neighboring jurisdictions.

Keywords: Local option tax, small business, spatial panel model

This is a draft, please do not cite without permission of the authors.

2

Local Fiscal Policies and Small Businesses in Florida:

A Spatial Panel Analysis

Introduction

The geography of new and small businesses, in light of their abilities to innovate and

create new jobs, has been extensively examined in terms of their geographic distribution, their

social, economic, and policy determinants within and across jurisdictions, and their impacts on

wealth and job creation over time and across geography. Consequently, a great number of

international, national, and local policies and programs have been established for fostering small

businesses (e.g., National Conference of State Legislatures, 2012; OECD, 1998). Among the

available studies on startup determinants, various governmental assistance and incentive

programs, such as small business loans or empowerment zones, have undoubtedly been one

important focus in part because findings of such studies can be directly and rapidly translated

into policy design, implementation, and evaluation processes in a continued effort to cultivate,

attract and retain new and small businesses.

The wide use of governmental incentives and concessions for alluring existing

manufacturing firms (smokestacks chasing) has been criticized as “zero sum games” when

anticipated benefits are actually smaller than offered various governmental concessions.

However, governmental incentives and concessions for alluring entrepreneurs and prospective

business owners in hopes of fostering entrepreneurship and small business development are

equally subject to the “zero sum games” criticism (Shane, 2009). Focusing simply and

sometimes exclusively on assistance and incentive programs help produce the illusion as if

entrepreneurs and small business owners are only interested in and are affected by such programs.

They are not. On the contrary, they are first and foremost influenced by economic fundamentals

3

and by policy decisions, which are very unlikely to be made with a preconceived notion of

entrepreneurship and small business assistance or development.

The purpose of this paper therefore is to examine the effects of local fiscal policies, for

example local option taxes, which appear in most cases not to be driven by intentions to develop

entrepreneurship and small businesses, on small business establishments. A case study of Florida

and its constituent counties will be conducted and presented primarily because of tax uniformity

of a single state and great tax discrepancies across states which make inter-state analyses much

difficult. A spatial panel econometric model will be calibrated to highlight potential temporal and

spatial effects of such local taxes and expenditures. An important implication of this study to

economic and entrepreneurship development policy makers and practitioners would be that there

is an alternative, probably more important, entrepreneurship development approach focusing on

essential local tax and expenditure structures and decisions as opposed to simply creating new

and costly incentives and concessions, which would nevertheless incur unnecessary bidding wars

and would exert minimum if any impacts on new and small businesses.

Literature Review

The creation of new, often small businesses are often regarded as an individual choice

(e.g., Knight, 1921) out of the trade-off among unemployment, self-employment or employment,

depending on the relative costs and returns of the three alternatives. For instance, opening a start-

up would be an entrepreneur’s rational choice if the perceived profits from owning a new firm

exceed his/her actual or perceived wage (Evans and Jovanovic 1989; Evans and Leighton 1989).

As a result, new firm formation is generally regarded as a behavioral manifestation of

entrepreneurship (Hebert and Link 1989; Wennekers and Thurik 1999), or an organizational

extension of individual entrepreneurial actions (Gartner 1989; Lumpkin and Dess 1996).

4

Guided by the individual choice approach, entrepreneurs’ idiosyncratic individual traits

and human capital characteristics that may be conducive to entrepreneurship spirit have been

extensively examined in the prior literature. Over the years, the list of individual characteristics

has grown and includes alertness, intelligence, aggressiveness, business leadership, high risk

tolerance, internal locus of control, motivation to succeed, previous experience, and so on

(Herbert and Link, 1983; Johnson and Cathcart, 1979; Mokry, 1988; Shaver and Scott, 1991;

Wiklund and Shepherd, 2003). Individuals who possess one, a few, or a combination of such

traits are regarded to be more likely to start and own new and small businesses. Such studies on

personal characteristics often rely on survey methods, which in most cases suffer from low

response rates. Further, the statistical relation between firm creation and growth and personal

traits and intentions tends to be rather weak. It is likely that the effect of individual intentions is

moderated by environmental or contextual factors, for example, access to resources.

In addition to the possession of individual traits, geographic concentration of population

with certain human capital characteristics was another important reason for explaining spatial

patterns of new firm formation across geography. Geographic concentration of human capital in

terms of either educational attainment (Acs and Armington, 2006; Bates, 1991; Evans and

Leighton, 1990) or the percentage of adults with college degrees (Glaeser, Scheinkman, and

Shleifer, 1995) has also been indicated as an important factor influencing the geography of

newly created businesses.

Earlier studies on existence of individual traits and geographic distribution of population

with certain human capital characteristics may be able to shed light on individual choices of

opening and owning new businesses and on uneven geographic distribution of new and small

ventures. However, such line of research has not fully addressed why individuals with

5

entrepreneurship friendly traits and human capital characteristics are geographically concentrated.

This paper relies explicitly on the Tiebout competition and focuses primarily on local tax and

expenditure structure on location choices of individuals’ potentially entrepreneurs and on

location choices of new firms created by individuals’ with entrepreneurship conducive traits and

human capital characteristics.

The optimal bundle of local taxes and public goods provided by local governments,

according to the Tiebout competition model (Tiebout, 1956), has played a critical role in

deciding where they would like to live. Tiebout posits that local jurisdictions, often within a

metropolitan area, provide their residents different combinations of public goods and government

services, such as parks and recreation, by imposing upon the residents different tax rates as

prices for public goods and services provided. Because residents tend to have varied preferences

and valuations of the goods and services and have varied willingness and ability to pay, residents

are hypothesized and actually observed to be able to “vote with feet.” In another words,

individuals, driven by utility maximization, will be able to shop around and even move from one

jurisdiction to another.

The Tiebout sorting process for residential location choices may be expanded to shed

light on the location decisions of new startups. First, for new startups, their locations are often

coupled with business owners’ residential location choices in light of the pervasiveness of home-

based small businesses in the United States. With the advent and surge of affordable information

technologies and e-commerce logistics, more and more entrepreneurs are launching businesses

from their homes. It is estimated in the 2012 Global Entrepreneurship Monitor Report that in the

United States over two-thirds of all new firms started at home, and about 59% of established

businesses with active employees continue to be operated out of business owners’ homes (Kelley

6

at al., 2012). The coupling of residential and firm location choices, i.e., where to start a business

equates where to live, renders entrepreneurs to the Tiebout sorting process in deciding the co-

locations of their homes and startups.

The second reason that the Tiebout sorting process for residential location choices is

highly relevant for startup locations is because startups are often geographically proximate to

entrepreneurs’ residence. The decision of where to open and operate a business is likely

dominated, if not pre-determined, by where to live even if new firms are founded and operated

outside of entrepreneurs’ residences (Baltzopoulos and Broström, 2011). Entrepreneurs tend to

create and run their new firms locally for tapping into local knowledge of relevant business

resources (Koster and Venhorst, 2014), for utilizing local social ties and networks (Dahl and

Sorenson, 2012), and for accessing localized business financing (Kerr and Nanda, 2009). In

addition, even in the absence of the benefits derived from locational match between residential

and firm locations, Stam (2007) suggested entrepreneurs’ residential location preferences and

behaviors may supersede firm interests when entrepreneurs may actively pursue and realize their

optimal residential choices even if this may incur additional costs for the firms. Figueiredo et al.

(2002) also showed that entrepreneurs and small business owners are willing to keep their

businesses in areas where they reside at the expense of facing and absorbing higher labor costs.

The highly aligned residential and firm location decisions make the location choices of small

businesses to be highly subject to the Tiebout sorting process.

Last but not least, even when residential and business locations are separated and

assumed to be independent, firm locations are influenced by local fiscal and institutional

situations which are essential to the Tiebout theory. Besides traditional cost and demand factors

(for a review, Isard, 1956), institutional and other non-economic factors have played increasingly

7

important roles in location decisions. Personal preferences on and hence personal satisfaction

from quality of life, social environment, weather, and landscape are significant determinants of

where to open and operate a firm (Hoover, 1948; Greenhut, 1967; Richardson, 1969).

Recognizing the critical role of personal factors, Tiebout (1957) stressed that small firms

compared to their larger counterparts would more likely be influenced by these non-pecuniary

determinants. Recent literature on creative class and talent attraction and retention further

emphasizes urban, cultural, and residential amenities and their role in attracting and cultivating a

local, productive workforce (Florida, 2002; 2004). The amenity-oriented firm location approach

maintains that amenities not only attract residents but also firms, and suggests business

entrepreneurs and owners in their location decision process will evaluate certain amenities with

respect to the likely residential locations of their employees (Koster and Venhorst, 2014; Gottlieb,

1995). In this sense, selecting business locations can be united with residential locations through

the availability and provision of residential amenities which are at the core of the Tiebout sorting

theory.

This paper is therefore built upon the Tiebout sorting process for residential location

choices and further expands it to test how entrepreneurs or prospective entrepreneurs sort

alternative locations for opening and running their new businesses when facing various bundles

of local public goods and services and taxes imposed. The Tiebout sorting of business

entrepreneurs and owners are not entirely different from that for general residents. Imposed taxes

are a cost factor both for the utility maximization of general residents and for profit

maximization of new and small businesses. Prospective and current entrepreneurs will simply

open start-ups in a community that offers a level of public goods and services that suits them.

The business location decisions however can be separated from for residential location choices

8

because the optimal bundles of goods and services and associated taxes may not be the same for

business profit maximization and for household utility maximization.

Research Question and Hypotheses

This paper is intended to examine how local taxes and expenditures would affect small

business establishments, under the assumption that business owners, like residents and

households, will also be involved in Tiebout sorting processes for finding locations that provide

public goods and services suited for their needs. Four major local taxes are empirically tested in

the paper, namely, property tax, local option sales tax, local option fuel tax, and local

communications service tax. Florida state constitution authorizes local government to collect

property tax for their operation, which accounts for 73% of total county tax revenue in 2013. The

local option sales tax in Florida is part of the eight types1 of surtax. Both the potential maximum

surtax rates and actual rates vary among county governments due to different combinations of

levies in each county. In addition, authorized by the state legislature, county governments in

Florida have three types of fuel tax levies that add up to 12 cents in total, but the actual rates vary

as some counties choose not to levy or not to levy to the maximum rate. County governments in

Florida can choose to levy local communication service tax by ordinance and set the rate at their

discretion (EDR, 2014). The total revenue yielded by these taxes takes up to 97% of the total tax

revenue of all counties in Florida in 20132. The four types of taxes selected are all tightly related

to tax authority of local jurisdictions but are not traditionally used for small business attraction

and development nor examined for their effects on locations of small business establishments. A

case study of Florida will be presented in the paper in part to maintain tax uniformity and

1 The eight surtaxes include charter county and regional transportation system surtax, local government

infrastructure surtax, small county surtax, indigent care/trauma center surtaxes, county public hospital surtax, voter-

approved indigent care surtax, emergency fire rescue services and facilities surtax, and school capital outlay surtax. 2 Authors’ calculation is based on the data provided by the Florida Department of Financial Services Bureau of

Local Government.

9

comparability across jurisdictions. In addition, this paper will only examine the manufacturing

sector (NAICS 31-33) and will be expanded in the future to include all industries for highlight

potential sectoral differences.

Three hypotheses will be tested in this paper, and they are:

Hypothesis 1: Higher millage rate of local property tax would hinder small business

creation.

Hypothesis 2: Higher local option sales tax rates, local option fuel tax rate, and

communication services tax rate hinder the creation of small business establishments.

Hypothesis 3: Higher local expenditure on physical environment, economic environment

and transportation encourage the creation of small business establishments.

Data and Variables

The areal units for the examination are Florida’s 66 individual counties3. The county-

level analysis is superior to most of prior studies because they were carried out at a highly

aggregated geographic level, for example, individual states (e.g., Carree 2002; Acs and

Armington 2006) or labor market areas (LMAs) (e.g., Armington and Acs 2002; Lee et al. 2004;

Reynolds et al. 1995). The earlier approaches may not only mask intra-unit variations, but also is

susceptible to modifiable areal unit problem (MAUP) bias when underlying economic relations

do not match with the boundaries of administratively determined states or other aggregated

geographic units.

For each of the counties, small business numbers are standardized by the size of the

county’s total labor force, in consistent with the labor force approach established in Audretsch

and Fritsch (1994). Information of small businesses in all Florida’s counties are obtained from

3 Jacksonville-Dual County is excluded from the analysis, because it is a consolidated city-county government.

10

the County of Business Pattern of the U.S Census Bureau. The publicly available information

measures primarily the stock of small businesses categorized by their numbers of employees, and

the dependent variable in spatial regressions is the stock of small businesses with 1-4 employees

in 2007-2012. It is important to note that the publicly available information based on stock of

businesses cannot provide accurate numbers of newly created firms over years, because the net

changes of the stocks of firms over years are influenced jointly by firm births and deaths during

the periods. In addition, small businesses are not synonymous with entrepreneurship because all

not small, new, and young businesses are driven by entrepreneurial spirit, but generally small

businesses have been widely used as an indicator for entrepreneurship.

Key individual variables consist of four types of local taxes, namely, property tax, local

option sales tax, local option fuel tax, and local option communications tax. In addition,

individual counties’ per capita expenditures in transportation, physical environment, and

economic environment are also collected and used in the regression analyses for controlling for

different usages of tax revenues across local jurisdictions. Both the local tax and expenditure

information is collected from Florida Department of Financial Services Bureau of Local

Government. Furthermore, commonly agreed determinants for small businesses are also included

in the regression models, and they are average employment per establishment for all industries,

share of manufacturing ventures in all establishments in a county, annual per capita income

growth rate, annual population growth rate, share of proprietors in labor force, and annual

population growth rate. Table 1 and 2 present detailed descriptions, data sources, and summary

for all the independent variables, and Table 3 presents a correlation matrix of all independent

variables used and no serious multicollinearity concern is suggested from the correlation matrix.

[Table 1, 2, 3 about here]

11

Spatial Panel Model

The longitudinal data of 66 counties ranging from 2008 to 2012 provide richer

information in examining how local government tax and expenditure affect small business

establishments. Moreover, the analysis also considers spatial relationship among counties. Not

only do the business establishments exhibit spatial dependence, local governments finance data

often has spatial autocorrelation, which is often neglected, Brueckner (2003) demonstrates the

necessity of spatial economic methods in the study of local governments’ public finance. For

example, in the quantile map shown in Figure 1, manufacturing small businesses in 2012 with 1-

4 employees normalized by total labor force are not evenly distributed, and instead, they are

geographically concentrated in the South Florida as well as both coastal lines. In addition, in the

quantile maps shown in Figure 2-5, the four major local taxes also suggest various degrees of

spatial correlation.

Traditionally, spatial econometric models consider the spatial dependence occurs either

in the dependent variable or the error term, which respectively lead to spatial autoregressive

(SAR) model and spatial error model (SEM). The SAR model is used under circumstance that

spatial autocorrelation exists in dependent variable. The SEM model considers the spatial

dependence exists in the error term. The OLS estimation of SEM model is consistent but no

longer efficient, which will lead to incorrect inference. Lesage and Pace (2009) advocates the

spatial Durbin model, which includes spatial lags of independent variables, because it can reduce

the omitted variable bias. The following equation represents different aforementioned spatial

models, when certain parameters are set to zero.

𝑌𝑖𝑡 = 𝛿 ∑ 𝑊𝑖𝑗

N

𝑗=1

𝑦𝑗𝑡 + 𝑥𝑖𝑡𝛽 + ∑ 𝑊𝑖𝑗

𝑁

𝑗=1

𝑥𝑖𝑗𝑡𝛾 + 𝜇𝑖 + 𝜆𝑡 + 𝑢𝑖𝑡

12

𝑢𝑖𝑡 = 𝜌 ∑ 𝑊𝑖𝑗

𝑁

𝑗=1

𝑢𝑖𝑡 + 휀𝑖𝑡

where i and j are the index for cross-sectional units and t is the time unit. Yit is the dependent

variable, i.e., stock of small business establishments in county i in year t. xit contains a matrix of

explanatory variables listed in Table 1. The error term 𝜇𝑖 and 𝜆𝑡 represent cross-section invariant

and time invariant unobserved effect. The inclusion of the spatial weighting matrix—W makes it

a spatial panel model. The weighting matrix W “expresses for each observation (row) those

locations (columns) that belong to its neighborhood set as nonzero elements” (Anselin and Bera,

1998, p. 243). As shown in the equation 1, the position of the weighting matrix indicates

different spatial econometric models. If the model only contains spatially lagged dependent

variable, i.e., 𝛾 and 𝜌 are set to zero, equation 1 represents for the SAR model. If the spatial

autoregressive process is in the error term, i.e., 𝛿 and 𝛾 are sent to zero, it stands for the SEM

model. With both spatially lagged dependent variable and explanatory variables on the right-

hand side of the equation (𝜌 = 0), the equation is for the SDM model.

In order to select the appropriate spatial model to conduct the analysis, we follow

Elhorst’s (2010) approach that consists of two steps—specific to general and general to specific.

Non-spatial model is estimated first and a Lagrange Multiplier (LM) test is performed to select

between a SAR and a SEM model. As long as the LM test indicates the existence of spatial

correlations, the general to specific step is performed. In the second step and from an opposite

direction, a SDM model is estimated and a Likelihood Ratio (LR) test will indicate whether the

SDM model can be simplified as a SAR or a SEM model. If the LM test of the first step supports

SAR model and the LR test, from an opposite direction, after the SDM confirms this (it can

reduced to SAR model), a SAR model is deemed appropriate. The same logic applies to the

13

selection of SEM model. However, if the LR test after the SDM model estimation shows that the

model cannot be simplified, or the test result contradicts with the previous LM test, the SDM

model will be preferred, because the SDM model yields consistent estimation4 regardless of the

spatial data generation process in the equation 1. In addition, all three models include cross-

sectional and time fixed effect (𝜇𝑖 𝑎𝑛𝑑 𝜆𝑡), in this case county and year fixed effect. The LR test

is utilized to test these fixed effects.

Empirical Findings

Table 4 presents the estimation results of the one-way fixed effect model that does not

consider the spatial autocorrelation. The LR test for joint significance of county fixed effects and

year fixed effect indicate the null hypothesis—𝐻𝑜:: 𝑢1, 𝑢12, … , 𝑢𝑛, = 0 should be rejected (LR:

458.94, df 66, P-value<0.0001), however the null hypothesis—𝐻𝑜:: 𝜆1, 𝜆12, … , 𝜆𝑛, = 0 cannot be

rejected (LR: 8.78, df 5, P-value= 0.12). Therefore, only county fixed effect is included in the

model.

[Table 4 about here]

In the one way fixed effect model, none of the variables in the category of tax rates

exhibits a statistically significant effect on the stock of small business establishments. The local

government’s per capita expenditure on economic development shows a significant and positive

effect on the small business establishments, which means that if a county government spends

more per capita on programs such as employment opportunity and development, industry

development, veteran’s services, and housing and urban development, it will create a business

friendly environment that attracts more entrepreneurship activities—more small business

establishments. All the six traditional factors have expected signs, among which three of them

4 Please refer to Elhorst (2010) for detailed discussion of the test procedure for spatial panel models.

14

are statistically significant, namely industry concentration, average establishment size, and

unemployment rate. Consistent with the earlier literature, the stock of small business

establishments tend to increase with higher industry concentration, smaller average establish size,

and higher unemployment rate. It appears that the non-spatial one-way fixed effect model only

explains 35 percent of the variations of the dependent variable.

Table 4 presents the estimation results of SDM. The LM tests after the non-spatial one-

way fixed effect model does not clearly point to either SAR or SEM model.5 However, as

suggested by Elhorst (2010), the general to specific approach is needed for the spatial model

selection. The LR test after the SDM model estimation indicates that the SDM model cannot be

simplified to neither SAR nor SEM model. Therefore, when suggestions from both directions are

inconsistent, the SDM model is selected, because SDM will produce unbiased estimation

regardless of the data generation process implied by equation 1 (Lesage & Pace, 2009). The

spatial lagged dependent variable and explanatory variables are represented by the multiplication

of a contiguity-weighing matrix for the counties in Florida.

[Table 4 about here]

The R-square of the calibrated SDM model is 0.86 due to the cross-sectional fixed effect

accounting for much of the variation. Elhorst (2010a) suggest using the difference between the

R-square and the squared correlation coefficient between actual and fitted value (corr2) to

indicate the proportion of variations that are explained by fixed effect. In this case, the corr2

equals to 0.40, which indicate that the fixed effects explain 46% of the variation of the dependent

variable.

5 LM test no spatial lag= 5.714, P-value 0.017; LM test no spatial error=3.547, P-value 0.06. Both null hypotheses

that there is no spatial lag and no spatial error are rejected; the robust LM tests do not yield clear indication as well.

Robust LM test no spatial lag=2.475 P-value 0.116; Robust LM test no spatial error=0.308 P-value 0.579.

15

LeSage and Pace (2009, p.74) indicates that the point estimates of the spatial models do

not yield proper marginal effects. In order to better capture the marginal effect and spatial spill-

over effect in a spatial model, they suggest a partial derivative interpretation for cross-sectional

cases. Elhorst (2010) extends this approach to the panel spatial models.6 The partial derivatives

of the dependent variable with respect to an explanatory variable at a time point consist of a

matrix with the same size of the weight matrix—W. This approach decomposes a total marginal

effect into a direct effect and an indirect effect. The direct effect is the average partial derivatives

of the diagonal elements of the matrix, which indicate how the change of an explanatory variable

in a county affects the dependent variable in the same county. The indirect effect is the average

derivative of the off-diagonal elements in the matrix and captures the spillover effect. In other

words, the indirect effect indicates how the change of an explanatory variable in a county affects,

on average, the dependent variables in neighboring counties. Both direct and indirect effects are

calculated and presented in Table 5.

[Table 5 about here]

After considering the spatial spillover effect, the estimation of the SDM model indicates

that county governments’ tax and expenditure policies play an important role in the development

of small business establishments. The property tax millage rate does not exhibit a statistically

significant direct effect on the small business establishments, but it has an indirect, spillover

effect, which is positive and significant at 10 percent level. This means that if the millage rate is

higher in a county, the stock of small business establishments will increase in its neighboring

counties, as if they are “pushed” into the nearby counties because of the “home” county’s higher

property tax rate. This implies that small business owners do take property tax burden into their

6 Please refer to Elhorst (2010, p.10) for detailed discussion and formulas.

16

consideration, and they are more likely to choose a jurisdiction for their new business

establishments with relatively low(er) millage rate. What matters here is not just the absolute

millage rate in a jurisdiction but how it is compared to neighboring counties.

The surtax rate also exhibits a significant indirect, spillover effect, but the negative

direction is against the expectation. It indicates that if a county imposes a higher sales surtax rate,

the stock small business establishments decreases in its neighboring jurisdictions. One possible

explanation for this unexpected result is that the local option sales surtax in Florida are levied by

county governments for different purposes, which include charter county transit system surtax,

local government infrastructure surtax, small county surtax, indigent care/trauma center surtaxes,

public hospital surtax, voter-approved indigent care surtax, and school capital outlay surtax.

Particularly, 31 counties are eligible to level small county surtax and 23 of them choose to do so,

which is the most frequently used local option sales surtax. This yield a possible explanation of

the negative indirect effect of the surtax rate: the counties with higher surtax rates tend to be

small counties, but business owners tend to choose bigger counties to establish their businesses.

The economic environment expenditure remains statistically significant. Both the direct

and indirect effects are positive, suggesting that if a county government spends more per capita

to improve the economic environment, the stock of small business establishments increases not

only in its own jurisdiction but also in the neighboring jurisdictions. In other words, the more the

economic environment expenditure in both its own jurisdiction and its neighboring jurisdictions

are, the more small business establishments increase. The business owners will compare the tax

rate among counties and prefer the one with lower rate. The economic environment expenditure

exhibits the spillover effect that exerts positive externality.

17

The traditional factors remain the expected signs after controlling the spatial

autocorrelation. Industry concentration has both positive direct and indirect effect, which

demonstrates the importance of industry concentration or specialization to small business

establishments. Average establishment size only exhibits negative direct effect at 10 percent

level. The unemployment rate has a positive direct effect and a negative sign of the indirect

effect though not significant. It suggests that the higher unemployment rate in a jurisdiction, the

more small business establishments in a jurisdiction. The effect of income growth exhibits the

same pattern as unemployment rate. The jurisdiction with higher income growth rate attracts

more small business establishments.

Conclusions

Entrepreneurship and small business establishment have played an essential role in the

economic development. Cultivating new business creation and high growth ventures have then

been an important task and goal for policy makers at various levels. However, the “zero sum

games” criticism against traditional smokestacks chasing economic development strategy not

only has given rise to the home grown entrepreneurship-led economic development paradigm,

but also has reminded us that the “zero sum games” criticism may be applied to excessive

governmental incentives intended for small businesses. As governmental incentives offered by

different jurisdictions are being cancelled out, the economic and policy fundamentals of a

jurisdiction are really matter to prospective entrepreneurs and business owners.

This research directly examines how small businesses would respond to local fiscal

policies, both taxes and expenditures, which generally are not made for entrepreneurship and

small business development. It expands the traditional Tiebout model, which focuses primarily

on residential location decisions, to explain how entrepreneurs would respond to taxes imposed

18

and public goods provided and how they would decide on the business locations that would

maximize their profits. It is hypothesized that a local jurisdiction’s fundamental fiscal

environment created by different bundle of taxes and expenditures affect the small business

owners’ decisions of their business locations. A case study on Florida counties and very small

manufacturing businesses (1-4 employees) is conducted and proper spatial panel econometric

model is calibrated in light of great spatial correlation. In accordance with Elhorst’s (2010)

procedure of selecting spatial models, the SDM model is supported after series of tests and yield

estimation results with least omitted variable bias. The SDM model results show that millage rate

has indirect, spillover effect on small business establishment, which demonstrates that small

business owners care about the property tax burden compared to neighboring counties. The per

capita expenditure of economic development programs exhibits spillover effect, which suggests

that small business owners prefer to locating their businesses in a county with higher expenditure

on economic development which surrounded by counties that have higher economic

development expenditure. This research shows that the overall fiscal environment to some extent

affects small business establishments.

This study uses data from only one state, which may limit its generalizability; however

one state data allows us to control state-specific heterogeneity and to take advantage of tax

uniformity. This study only has five-year data that cover the cycle of the recent Great Recession.

A longer panel may be more informative in the way that local governments’ long-term fiscal

environment can be captured. The future study may also consider small business establishments

in other sectors and breakdown of local option sales surtax.

19

References

Acs, Z. and Armington, C. (2006). Entrepreneurship, geography, and American economic

growth. Cambridge, MA: Cambridge University Press.

Armington, C. and Acs, Z. (2002). The determinants of regional variation in new firm formation.

Regional Studies, 36, 33-45.

Audretsch, D. and Fritsch, M. (1994). On the measurement of entry rates. Empirica, 21, 105-113.

Baltzopoulos, A. and Broström, A. (2011). Attractors of entrepreneurial activity: universities,

regions and alumni entrepreneurs. Regional Studies, 47, 934-949.

Bates, T. (1991). Commercial bank financing of white and black owned small business start-ups.

Quarterly Review of Economics and Business, 13, 64-80.

Carree, M. (2002). Does unemployment affect the number of establishments? A regional analysis

for US states. Regional Studies, 36, 389-398.

Dahl, M. S. and Sorenson, O. (2012). Home sweet home: entrepreneurs' location choices and the

performance of their ventures. Management Science, 58, 1059-1071.

Evans, D. and Jovanovic, B. (1989). An estimated model of entrepreneurial choice under

liquidity constraints. Journal of Political Economy, 97, 808-827.

Evans, D. and Leighton, L. (1989). Some empirical aspects of entrepreneurship. American

Economic Review, 79, 519-535.

Figueiredo, O., Guimarães, P., and Woodward, D. (2002). Home-field advantage: Location

decisions of Portuguese entrepreneurs. Journal of Urban Economics, 52, 341-361.

Florida, R. (2002). The rise of the creative class. New York: Basic Books.

Florida, R. (2004). Cities and the creative class. New York: Routledge.

Florida Legislature’s Office of Economic and Demographic Research. (2014). 2014 Local

Government Financial Information Handbook. Retrieved from the Office of Economic

and Demographic Research website: http://edr.state.fl.us/Content/local-

government/reports/lgfih14.pdf.

Gartner, W. B. (1989). “Who is an Entrepreneur?” is the Wrong Question. Entrepreneurship

Theory and Practice, 13, 47-68.

Glaeser, E., Scheinkman, J., and Shleifer, A. (1995). Economic growth in a cross-section of

cities. Journal of Monetary Economics, 36, 117-143.

Gottlieb, P. D. (1995). Residential amenities, firm location and economic development. Urban

Studies, 32, 1413-1436.

Greenhut, M. L. (1967). Plant location in theory and in practice. Chapel Hill: University of

North Carolina Press.

Hebert, R. F. and Link, A. (1983). The entrepreneur: Mainstream views and radical critiques.

Southern Economic Journal, 50(2), 611-612.

Hebert, R. F. and Link, A. (1989). In search of the meaning of entrepreneurship. Small Business

Economics, 1(1), 39-49.

Hoover, E. M. (1948). Location of economic activity. New York: McGraw-Hill.

Isard, W. (1956). Location and space-economy: A general theory relating to industrial location,

market areas, land use, trade, and urban structure. Cambridge, MA: MIT Press.

Johnson, P. and Cathcart, D. (1979). The founders of new manufacturing firms: A note on the

size of incubator plants. Journal of Industrial Economics, 28, 219-224.

Kelley, D., Ali, A., Brush, C., Corbett, A., Majbouri, M., and Rogoff, E. (2012). Global

entrepreneurship report: National entrepreneurial assessment for the United States of

20

America. Retrieved on April 25, 2016, from the website,

http://www.babson.edu/Academics/centers/blank-center/global-

research/gem/Documents/GEM%20US%202012%20Report%20FINAL.pdf.

Kerr, W. and Nanda, R. (2009). Democratizing entry: Banking deregulations, financing

constraints, and entrepreneurship. Journal of Financial Economics, 94(1), 124-149.

Knight, F. (1921). Risk, uncertainty and profit. New York: Houghton Mifflin.

Koster, S. and Venhorst, V. (2014). Moving shop: residential and business relocation by the

highly educated self-employed. Spatial Economic Analysis, 9(4), 436-464.

LeSage, J. and Pace, R. K. (2008). Introduction to spatial econometrics. Boca Raton, FL:

Chapman and Hall.

Lee, S., Florida, R., and Acs, Z. (2004). Creativity and entrepreneurship: A regional analysis of

new firm formation. Regional Studies, 38, 879-891.

Lumpkin, G. and Dess, G. (1996). Clarifying the entrepreneurial orientation construct and

linking it to performance. Academy of Management Review, 21, 135-172.

Mokry, B. (1988). Entrepreneurship and public policy: Can government stimulate business

startups? New York: Quorum Books.

National Conference of State Legislatures (2012). Promoting entrepreneurship: Innovations in

state policy. Retrieved from website:

http://www.ncsl.org/documents/fiscal/entrepreneurshipFINAL05.pdf.

OECD (1998). Fostering Entrepreneurship: The OECD Jobs Strategy. Paris, OECD.

Reynolds, P., Maki, B., and Miller, W. (1995). Explaining regional variation in business births

and deaths: US 1976-88. Small Business Economics, 7: 389-407.

Richardson, H. W. (1969). Regional economics: Location, theory, urban structure, and regional

change. New York: Praeger Publishers.

Shane, S. (2009). Why encouraging more people to become entrepreneurs is bad public policy.

Small Business Economics, 33, 141-149.

Shaver, K. and Scott, L. (1991). Person, process, choice: The psychology of new venture

creation. Entrepreneurship: Theory and Practice, 16(2), 23-45.

Stam, E. (2007). Why butterflies don’t leave: Locational behaviour of entrepreneurial firms.

Economic Geography, 83, 27-50.

Tiebout, C. (1956). A pure theory of local expenditures. Journal of Political Economy, 64, 416-

424.

Tiebout, C. (1957). Location theory, empirical evidence and economic evolution. Papers and

Proceedings of the Regional Science Association, 3, 74-86.

Wennekers, S. and Thurik, R. (1999). Linking entrepreneurship and economic growth. Small

Business Economics, 13(1), 27-55.

Wiklund, J. and Shepherd, D. (2003). Aspiring for, and achieving growth: The moderating role

of resources and opportunities. Journal of Management Studies, 40(8), 1919-1941.

21

Table 1: Variable Descriptions and Sources

Variables Descriptions Expected

signs

Sources

Dependent variable

Small businesses Numbers of manufacturing small ventures (1-4 employees)

normalized by total labor force

County Business

Patten

Independent variables

Tax Rates

Florida Department

of Financial

Services Bureau of

Local Government

Millage rate Property millage tax rate –

Fuel tax rate Local motor fuel option tax –

Communication tax rate Local communication services tax –

Sales surtax rate Local option sales tax rate –

Expenditures

Transportation exp Transportation expenditure per capita +

Physical environment exp Physical environment expenditure per capita +

Economic environment exp Economic environment expenditure per capita +

Traditional factors

Industry concentration Share of manufacturing ventures in all establishments in a

county

+

U.S. Census Bureau Avg. establishment size Average employment per establishment for all industries –

Unemployment rate Unemployment rate +

Population growth Annual population growth rate

Income growth Annual per capita income growth rate +

Share of proprietors Share of proprietors in labor force + U.S. Bureau of

Economic Analysis

22

Table 2: Descriptive Statistics (N=330)

Mean

Std.

Dev. Min Max

Dependent Variable

Small businesses 47.80 14.59 0.00 80.77

Tax Rates

Millage rate 6.52 2.07 0.20 10.00

Fuel tax rate 49.93 2.57 44.90 54.60

Communication services tax rate 2.57 1.55 0.29 7.38

Surtax rate 0.73 0.45 0.00 1.50

Expenditures

Transportation expenditure 236.31 164.72 57.11 1718.20

Physical environment expenditure 178.28 -163.02 92.82 1278.45

Economic environment expenditure 57.49 -60.97 10.25 364.75

Traditional factors

Industry concentration 0.03 0.01 0.00 0.09

Average establishment size 10.67 2.47 6.13 19.74

Proprietors % 0.26 0.08 0.11 0.54

Unemployment rate 9.17 2.30 4.00 15.10

Population growth rate 0.01 0.01 -0.05 0.05

Income growth rate 0.01 0.04 -0.15 0.13

23

Table 3: Correlation Table of Independent Variables

M F CST S TE PEE EEE IC AES P% Unem PGR IGR

Millage (M) 1.000

Fuel tax rate (F) -0.175 1.000

Communication service tax rate (CST) -0.359 0.254 1.000

Surtax rate (S) 0.456 -0.363 -0.221 1.000

Transportation expenditure (TE) 0.050 -0.077 -0.177 0.076 1.000

Physical environment exp. (PEE) -0.380 0.286 0.078 -0.326 0.220 1.000

Economic environment exp. (EEE) -0.278 -0.091 -0.058 0.009 0.312 0.258 1.000

Industry concentration (IC) 0.440 -0.121 -0.137 0.181 0.045 -0.153 -0.179 1.000

Average establishment size (AES) -0.102 0.090 0.406 -0.157 -0.232 -0.065 0.041 -0.061 1.000

Proprietors % (P%) 0.181 -0.081 -0.241 0.141 0.261 0.074 0.011 0.230 -0.633 1.000

Unemployment rate (Unem) -0.043 0.071 0.010 -0.096 -0.101 0.058 -0.222 0.015 -0.010 -0.022 1.000

Population growth rate (PGR) -0.215 -0.001 0.129 -0.165 0.110 0.037 0.090 -0.076 0.153 -0.189 -0.225 1.000

Income growth rate (IGR) 0.099 -0.059 -0.090 0.048 -0.009 -0.123 0.017 -0.014 -0.012 -0.009 0.084 -0.060 1.000

24

Table 4: Non-spatial and Spatial Model Results

Dependent Variable: Number of Small Businesses Normalized by Total Estalishments

Non-Spatial OLS Spatial Durbin Model

Coefficients S.E. Coefficients S.E.

Millage -0.293 0.830 -0.533 0.928

Fuel tax rate -0.241 0.587 -0.013 0.667

Communication service tax rate 0.694 1.055 1.066 1.159

Sales surtax rate -4.933 4.427 -3.595 4.775

Transportation expenditure 0.002 0.003 0.004 0.004

Physical environment expenditure 0.003 0.005 0.005 0.006

Economic environment expenditure 0.047*** 0.014 0.054*** 0.001

Industry concentration 984.445*** 92.552 1031.467*** 102.3

Average establishment size -1.214* 0.705 -1.399* 0.797

Proprietors % -21.425 29.153 -23.348 36.799

Unemployment rate 1.205*** 0.314 1.812* 1.007

Population growth rate -7.625 36.389 -1.374 41.089

Income growth rate 20.982 8.962 34.621** 14.992

W*Dependent 0.063 0.068

W*Millage 2.498* 1.503

W*Fuel tax rate -0.420 1.185

W*Communication service tax rate -1.021 1.927

W*Sales surtax rate -15.099* 8.684

W*Transportation expenditure 0.000 0.006

W*Physical environment expenditure 0.004 0.012

W*Economic environment expenditure 0.070** 0.030

W*Industry concentration 511.098** 213.100

W*Average establishment size -0.479 1.171

W*Proprietors % 60.191 48.969

W*Unemployment rate -0.574 1.095

W*Population growth rate 32.231 64.599

W*Income growth rate -24.721 18.073

Summary Statistics

No. of observations 330 330

R-square 0.35 0.86

Log likelihood -1041.9

LM spatial lag 5.714**

Robust LM spatial lag 2.475

LM spatial error 3.547*

Robust LM spatial error 0.3076

Wald spatial lag 18.34

LR spatial lag 21.52*

Wald spatial error 20.88*

LR spatial error 24.89**

25

Table 5. Spatial Durbin Model (SDM) Decomposition Results

Dependent Variable: Number of Small Businesses Normalized by Total Establishments

Direct Effect Indirect Effect Total Effect

Tax rates

Millage

-0.524

(0.948)

2.66*

(1.597)

2.136

(1.904)

Fuel tax rate 0.012

(0.673)

-0.469

(1.221)

-0.456

(1.260)

Communication

service tax rate

1.090

(1.142)

-0.945

(1.944)

0.145

(1.260)

Surtax rate -3.864

(4.806)

-15.688*

(8.900)

-19.551*

(10.32)

Expenditures

Transportation expenditure

0.004

(0.004)

0.00002

(0.007)

0.004

(0.008)

Physical environment

expenditure

0.005

(0.006)

0.004

(0.012)

0.009

(0.014)

Economic environment

expenditure

0.056***

(0.016)

0.077**

(0.03)

0.133***

(0.036)

Traditional factors

Industry concentration

1042.58***

(103.939)

606.16***

(215.914)

1648.735***

(252.542)

Average establishment size -1.414*

(0.756)

-0.545

(1.212)

-1.960

(1.395)

Proprietors % -22.964

(36.12)

59.743

(48.83)

36.779

(51.360)

Unemployment rate 1.799*

(0.983)

-0.457

(1.063)

1.342**

(0.573)

Population growth rate -0.270

(39.437)

38.640

(65.338)

38.370

(69.265)

Income growth rate 33.901**

(14.263)

-23.394

(18.222)

10.507

(13.046)

R-Square: 0.8332, Corr2: 0.2149

Log Likelihood: -1032.28

26

27

28

29

30