Академический Документы
Профессиональный Документы
Культура Документы
, (2)
where E is total operating expenditures per pupil;
(3)
In a typical school districts annual budget process, the district calculates per pupil
property tax revenues that need to be raised () by subtracting F and O from E. As the final step
to establish how much the district will spend (Mikesell 2014, p. 495), the property tax rate, , is
set by equation (4):
(4)
The first term of (4),
.
9
Substituting (4) into (1) yields
9
Total school district property tax rates were always higher than the uniform rate during the sample
period, ranging between $8.168 and $24.426 per $1,000 assessed valuation.
11
(
(5)
I now incorporate the potential effects of TIF into equation (5). Of the factors in this
equation, TIF is mostly likely to affect
), the TIF area has two separate taxable property values per pupil:
frozen value (
). While
produces additional revenue that, rather than being channeled to the school district, is captured
by the initiating city or county to finance TIF activities, the school district is still entitled to
and
, or
.
11
Assuming that there is no spillover effect of TIF on non-TIF areas (or on
), I focus first
on how
and thus
might be smaller or larger with TIF than without TIF. One of two common
10
Potential effects of TIF on education expenditures can be explained through a budgeting mechanism.
This mechanism also examines how TIF might affect
. The
budgeting mechanism is less appropriate for this study. While TIF might change
, a
key measure of TIF use, raises the tax liability of a voter (via
.
12
The first scenario illustrated by Figure 1 occurs when
captures the
difference between
and
. This is the best scenario in which local residents, all else being
equal, benefit from lower tax liabilities, thereby probably increasing their demand for education
spending.
13
Three points are worth mentioning. First, either scenario 1 or 2 may occur when
as in Figure 1, that is, when the values of properties within TIF districts have a higher rate of
appreciation than those of similar properties outside TIF districts (as in the case of Smiths
[2006] finding). Second, scenario 1 may even occur when the values of TIF properties trend
12
The term underlying indicates that the growth trend of
). Any increase in
can be a function of
. The value of
has a positive
(negative) underlying growth rate.
14
It is thus difficult to accurately predict whether the
relationship between TIF and Iowa school districts education expenditures is neutral, positive,
or negative. Instead, the data can inform us about this relationship.
The earlier review of the TIF literature also suggests that the spillover effects from TIF
may vary depending on the type of TIF district (i.e., commercial, industrial, residential). Put
differently, the size of the spillover effects and thus of the demand for education spending
(holding everything else constant) may vary depending on type of TIF property. The second
research question is intended to explore these potentially differential effects of different TIF
property types.
The above discussions apply to TIF districts during their life spans. When a TIF district
expires, its incremental value (
and what
, (6)
where i and t index school districts and years. In equation (6), represents district fixed effects
to control for time-invariant district-specific unobserved factors. As adopted in Downes and
Pogue (1994), these district fixed effects also account for time-invariant unobserved efficiency
factors, thereby addressing most of the relative efficiency differences across districts.
18
Year
dummies, , control for common factors affecting all school districts in Iowa in a given year.
Vector X includes all other control variables (unless otherwise specified, all financial
measures are in 2010 dollars and per pupil terms). The first group of control variables includes
characteristics of the student body, such as enrollment, percent of African-American students,
and shares of disadvantaged students, including those receiving free or reduced-price lunch, and
those with Limited English Proficiency (LEP). A higher concentration of disadvantaged students
is expected to induce greater expenditures. Studies reviewed in Fox (1981) and Andrews,
Duncombe, and Yinger (2002) provide empirical evidence that economies of size in education
may help larger districts incur lower costs per pupil. Following this literature, I specify the log of
per pupil expenditures as a quadratic function of the log of enrollment. The second group of
control variables represents voter income (measured by median household income following the
median voter framework of public choice to be discussed), income surtaxes, and
intergovernmental aid (namely, state and federal aid) that school districts receive.
19
Higher voter
18
Studies such as Duncombe and Yinger (2011) may specify efficiency factors that do change over time
as a function of voter income and tax price. However, I use different measures of tax price for vector Y,
and variables indicating voter income are included in vector X of equation (6). For ease of interpretation,
these efficiency effects are disregarded.
19
I do not have data on investment income or student fees. However, they are expected to play a
negligible role in counteracting the use of TIF. Both investment income and student fees traditionally
16
income and additional revenues from income surtaxes or intergovernmental aid are expected to
increase education expenditures.
The third group of variables includes variables representing district-level differences in
demographic characteristics, namely, the share of the population that graduated from four-year
college and the share of owner-occupied housing units. These populations, compared to non-
college graduates and renters, respectively, may have different preferences for school
expenditures. Specifically, college-educated people may desire greater school spending
(Bergstrom, Rubinfeld, and Shapiro 1982; Hilber and Mayer 2009). Since school quality is found
to be capitalized into property values (Nguyen-Hoang and Yinger 2011), homeowners are also
expected to prefer higher education expenditures.
20
<B>Variables of TIF Use
I have two sets of TIF-related measures (Y). I show in the theoretical framework that the
potential effects of TIF on education expenditures revolve mostly around the property tax base,
and
replace undefined logged values (when
) with 0.)
. Higher
and what
. As
previously discussed, the coefficient of
) in the
current year. This measure might also capture the same potential effects of TIF on
as
. The
coefficients of
and
and
are both
22
For school districts with multiple TIF areas, the estimated coefficient indicates the average effect of all
TIF areas.
18
estimated, the coefficient of
.
In equation (6),
),
24
and the number of expired TIF districts within a school district (
).
25
These two
variables are expected to be positively correlated with education expenditures.
To investigate the fourth question as to whether TIF affects the education expenditures
of lower-wealth school districts more than those of higher-wealth districts, I include in equation
(6) an interaction between the main measure of TIF (
or
.
24
This ratio captures the intensity of the TIF use in year
t1
. I use this measure instead of the incremental
taxable property value per pupil in the preceding year because I do not have enrollment data for the
academic year 19992000. The estimation results reported in Table 5 are still insignificant when I divide
the incremental property value in year
t1
by the enrollment in the current year, or when I log this ratio.
25
New TIF districts are in the current year but not in the immediately preceding year, while expired TIF
districts exist in the immediately preceding year but not in the current year.
19
districts relative level of wealth or income (W).
26
This variable, W, is coded 0, 1, and 2 for
districts in which the median housing value in Census 2010 was below the 33rd percentile,
between the 33rd and 67th percentiles, and above the 67th percentile, respectively. If the
coefficient of
and
. (The results
are almost identical when they are estimated separately.) The coefficient of
is negative but
insignificant, showing ambiguous evidence of this variable on education spending. The other two
TIF measures, namely, tax price (
),
show negative price effects on school expenditures, indicating lower demand for spending in
response to higher prices. A 1 percent increase in
and
by approximately one
standard deviation to slightly over $28,000. This 160 percent increase in
would induce a
reduction of only 0.29 percent (160 0.0018 percent) in this districts operating spending per
pupil ($31), thus, a reduction of $41,664 in total spending.
Many of the other variables are highly significant with expected signs. I find a quadratic
relationship between enrollment and education spending. While more students decrease
education expenditures per pupil, diseconomies of size start to set in at an enrollment of about
24,300. (In Iowa, only the enrollment of the Des Moines School District exceeds this threshold.)
A 1 percent increase in the share of low-income students is associated with an increase of nearly
0.02 percent in operating expenditures. Voters higher income also increases their demand for
education expenditures. The demand also becomes higher with increased income surtaxes and
state and federal aid. The effect of state aid is more than nine times larger than that of federal aid.
22
As with Nguyen-Hoang (2013), an increase in the shares of college graduates and homeowners
within a district is associated with higher school spending.
Table 4 reports the estimation results for the second research question regarding the
potential differential effects of different TIF types on education expenditures. This table shows
that incremental property values of residential and industrial properties in TIF districts are
negatively related with school district spending. An increase in the incremental values of
residential and industrial properties is associated with a reduction of 0.0012 percent,
respectively, in educational spending. This price effect of residential TIF on demand for
education spending can be explained by, as reviewed in Nguyen (2005), the negative, though
small, spillover effects of affordable housing on the values of nearby properties. (The primary
purpose of residential TIF in Iowa is to create affordable housing.) In keeping with the
theoretical framework, the negative price effect of industrial TIF on education expenditures may
be explained by negative spillover effects on nearby properties (Dye and Merriman 2003; Weber,
Bhatta, and Merriman 2007) and/or a positive growth trend of industrial TIF properties (Weber,
Bhatta, and Merriman 2003). The coefficient of commercial TIF is insignificant probably
because of mixed spillover effects of all commercial TIF districts within a school district on
nearby properties: some commercial TIF districts have positive spillovers (Merriman, Skidmore,
and Kashian 2011), others negative (Weber, Bhatta, and Merriman 2007).
Table 5 reports the results for the third and fourth research questions. In columns 1 and 2,
I find little empirical evidence for effects of expired TIF districts measured by either the ratio of
incremental taxable property value to total taxable value in year
t1
(
, changes in
) 0.12
(5.67)***
Log of incremental taxable property value
per pupil (
)
0.0018
(2.86)***
Number of new TIF districts within
school districts (
)
0.00033
(0.54)
Log of enrollment 1.01
(10.49)***
Squared log of enrollment 0.050
(7.40)***
Logged percent of free and reduced-price
lunch students
0.018
(2.12)**
Logged percent of LEP students
0.00036
(0.23)
Logged percent of African American
students
0.0017
(0.91)
Log of median household income
0.061
(1.74)*
Log of state aid per pupil 0.14
(4.28)***
Log of federal aid per pupil 0.015
(3.14)***
Log of income surtaxes per pupil 0.0028
(2.78)***
Percent of college graduates 0.0016
(1.74)*
Percent of owner-housing units 0.00077
(1.66)*
R
2
0.621
Notes: There are 3,817 observations between 2001 and 2011.
Regressions are estimated with district fixed effects and year
dummies. Hypothesis testing is done with the robust
heteroskedasticity and autocorrelation (HAC) Newey-West standard
errors. Numbers in parentheses are t-statistics.
* p < 0.10, ** p < 0.05, *** p < 0.01
34
Table 4 Effects of Different Property TIF Types
Key Variables Coefficients
Log of tax price (T
P
) 0.12
(5.66)***
Log of incremental per pupil taxable property value of residential property 0.0012
(1.85)*
Log of incremental per pupil taxable property value of commercial property 0.00036
(0.60)
Log of incremental per pupil taxable property value of industrial property 0.0012
(2.07)**
Number of new TIF districts within school districts (
) 0.00034
(0.55)
Notes: All of the unreported variables and the other notes are the same as in Table 3.
* p < 0.10, ** p < 0.05, *** p < 0.01
35
Table 5 TIF Effects of Expired TIF Districts and for Low-Income School Districts
Key Variables
Key Distinguishing Specification
Expired TIF
ratio (
and # of
expired TIF
districts (
)
School district
types based on
median housing
value
School district
types based on
median household
income
(1) (2) (3) (4)
Log of tax price (
)
0.0018 0.0018 0.0034 0.0038
(2.86)*** (3.21)*** (3.46)*** (3.79)***
Number of new TIF districts
within a school district (
)
0.00034 0.00034 0.00035 0.00041
(0.54) (0.54) (0.55) (0.64)
Ratio of incremental taxable
property value to total taxable
value in year
t1
(
)
0.020 0.019 0.023 0.022
(0.11) (0.09) (0.11) (0.10)
Number of expired TIF districts
within a school district (
)
0.000022 0.000038 0.000058
(0.01) (0.02) (0.03)
Logged
W 0.0019 0.0021
(2.49)** (2.81)***
Notes: All of the unreported variables and the other notes are the same as in Table 3. W in column 3 is
coded 0, 1, and 2 for school districts whose median housing values in Census 2010 were below the 33rd
percentile, between the 33rd and 67th percentiles, and above the 67th percentile, respectively. W follows
the same coding procedure, except that the Census 2010 median household income is used.
* p < 0.10, ** p < 0.05, *** p < 0.01