Some
tables and equations in this HTML document are under construction.
This
document in its complete form is available in PDF format
Executive
Summary
The
Texas State Tax Analysis Modeling Program
(Texas-STAMP)
Metodology and Applications
Prepared
by the Beacon Hill Institute
David G. Tuerck, PhD, Project Director
Jonathan Haughton, PhD, Project Manager
In-Mee Baek, PhD, Project Consultant
James Connolly, BA, Research Assistant
Scott Fontaine, BA, Research Assistant
This
study was commissioned by the Texas Public Policy Foundation and
conducted by The Beacon Hill Institute for Public Policy Research
at Suffolk University. ©1999 All rights reserved.
January
1999
Executive
Summary
For
many years there was a tradition in state government to treat taxes
as if they didn't matter. It was presumed that businesses and individuals
cared mainly about state and local government services and infrastructure:
States could raise taxes without regard to any adverse economic
effects.
That
tradition is giving way to an economic reality: State taxes do matter.
This is especially true in today's competitive economic environment
in which firms are able and willing to move from one state to another
as tax incentives become available to them.
Assessing
the effects of proposed tax-law changes using reliable, credible
methods is therefore crucial. In response to this demonstrated need,
the Beacon Hill Institute at Suffolk University in Boston developed
STAMP, the State Tax Analysis Modeling Program. STAMP permits policymakers,
opinion leaders and scholars to determine how a change in state
tax law will affect the state economy.
As
an “econometric model,” STAMP applies economic and statistical
methods to state and federal data to determine how state economic
indicators vary with changes in state tax law. The economic indicators
usually considered by STAMP are the number of jobs, wage rates,
the capital stock and tax revenues. The capital stock is the total
value of nonresidential capital, including manufacturing facilities,
office and warehouse space, business equipment and other items that
business uses in production.
STAMP
permits its users to “simulate” the effects on these indicators
of alternative tax-law changes and combinations of tax-law changes.
Using STAMP, we can, for example, estimate the effects on employment
of a simultaneous reduction in property taxes and increase in the
state sales tax.
States
frequently rely on “static” models that attempt to determine
the effects of tax-law changes on tax revenues. These models make
the simplifying assumption that changes in tax law have no effect
on economic activity. For example, in a static model, a rise in
the sales tax rate would be assumed to have no effect on sales.
A 20% rise in the sales tax rate would be expected to bring about
a 20% rise in sales tax revenue.
STAMP
is, in contrast, a “dynamic” model. Dynamic models determine
the effects of tax-law changes on tax revenues by taking into account
how those changes affect economic activity. In a dynamic model,
a rise in the sales tax rate would ordinarily be expected to bring
about a decrease in sales. Thus a 20% rise in the sales tax rate
would bring about a less-than-20% rise in sales tax revenue.
History
of STAMP
BHI
developed STAMP in 1994 in recognition of a need for a dynamic modeling
capability in Massachusetts. The Massachusetts STAMP was first applied
to a state ballot initiative that would have replaced the existing
flat tax with a graduated income tax. STAMP showed that the proposed
graduated income tax would have destroyed about 80,000 jobs and
more than $1 billion in wages. The voters overwhelmingly rejected
the initiative.
Subsequently,
at the request of the Oklahoma Office of State Finance, BHI developed
a STAMP for Oklahoma. In 1997, BHI provided evidence supporting
the view held by state officials that a cut in income or sales taxes
would increase personal income. Also in 1997, BHI built a New Jersey
STAMP that showed how recent income tax cuts had resulted in the
creation of about 25,000 new jobs and of more than $2 billion in
new capital spending.
In
January 1998, BHI released the results of an Ohio STAMP showing
how a proposed increase in the state sales tax from 5% to 6% would
destroy about 100,000 jobs and $3 billion in wages. The governor
had proposed the tax hike as a way of raising $1 billion in new
revenue to fund education spending. The state legislature rejected
the tax hike, as did the voters overwhelmingly in a subsequent referendum.
Other
tax analyses performed by BHI include an evaluation, in 1997, of
a proposed reduction in the Iowa income tax and, in 1998, of a proposed
oil processing tax in Louisiana. The Iowa study figured in a 10%
tax cut adopted by the Iowa legislature. The Louisiana study showed
that replacing the existing severance tax with a processing tax
would slow the state economy by raising refiners' costs and the
price of gasoline.
Texas-STAMP
Texas-STAMP
is STAMP applied to the Texas economy. Estimated from Texas and
U.S. data for 1970-1996 and for eight sectors of the Texas economy,
it is designed to trace the economic effects of Texas tax-law changes.
It can be applied to changes in the state sales tax, state unemployment
insurance and workers compensation tax rates, property tax rates,
and the state franchise tax. It shows how increases in these tax
rates exert measurable, negative effects on the state economy.
How
does STAMP work?
STAMP
works by using certain “coefficients” obtained from the
data to show how economic indicators such as jobs vary with changes
in tax law. The coefficients are obtained from the data by estimating
regression equations that link jobs and the other economic indicators
to factors, notably tax rates, that influence those indicators.
Economists
sometimes estimate the coefficients that link policy changes to
economic changes without specifying the underlying theory on which
those coefficients are based. STAMP avoids this error. The regression
equations in STAMP are “reduced-form” equations, derived
from a set of “structural equations” that lay out explicitly
the theoretical underpinnings of the model.
STAMP's
theoretical underpinnings are the stuff of standard economics texts:
Households demand goods and services. If government raises taxes
on the purchase of these goods and services, households will buy
less of them. Households also supply labor to firms. If labor services
are taxed more heavily, then households will supply less of them.
Firms
are assumed to maximize profit, which they do by combining labor
and capital in order to produce the goods and services that households
demand. In this process, firms supply goods and services. If they
have to pay a higher sales tax, then they will charge more for their
goods, sell less as a result, and have to reduce production. Firms
also pay wages and salaries to their workers. If the cost of hiring
their workers rises, for example, because unemployment insurance
payments grow, then they will economize on their use of labor by
cutting back or finding substitutes.
What
are the model's “coefficients?”
Our
estimates of the model's reduced form equations are as follows:
Equation
1: Jobs
| ln(number
of jobs) |
= |
-0.1342
ln(government transfers per nonworking adult)
(0.00) |
| |
|
-0.0250
Nonwage labor costs as a percent of the wage
(0.00) |
| |
|
+0.0061
Federal tax rate on labor income
(0.00) |
| |
|
-0.0077
Effective sales tax rate
(0.09) |
| |
|
-0.0306
Cost of capital
(0.00) |
| |
|
+0.6492
ln(working age population)
(0.00) |
| |
|
+
constant terms for each of eight sectors of the economy |
| |
|
+
terms with indices of U.S. output, for each of eight sectors
of the economy. |
There
are six coefficients, each with an associated p-value given below
it in parentheses. The p-values measure the statistical significance
of the coefficients. We follow the convention whereby a coefficient
is deemed statistically significant if its p-value is less than
.10. A p-value below 0.10 indicates that there is less than a 10%
chance that the true value of the coefficient is zero.
Here
the dependent variable is the logarithm of the number of jobs. The
first independent variable is the logarithm of government transfers
per nonworking adult. The coefficient for this variable reflects
the theoretical idea that transfer payments in the form of unemployment
and welfare benefits decrease the number of jobs. The negative coefficient
indicates that when these payments are larger, employment will be
lower, presumably because the transfers will mean that people have
less pressure to search for work. Specifically, the coefficient
implies that a 1% increase in government transfers decreases jobs
by .1342%.
Nonwage
labor costs refer to disability and unemployment insurance payments
made by employers. These costs have much the same effect as a tax
on labor, and the estimated equation shows that when the costs rise,
employers will cut back on hiring.
The
estimated equation also shows that higher property taxes or sales
taxes will reduce employment. The sales tax, as measured here, includes
the ordinary sales and use tax and also the state taxes on gasoline,
diesel fuel and motor vehicle sales.
A
higher cost of capital has the same effect. The cost of capital
measures how much employers have to pay to "hire" the
capital they use. It is a complex measure, which reflects the size
of depreciation, the cost of borrowing (i.e. interest rate), and
the numerous taxes that bear on capital (including the state corporate
franchise tax, the federal corporate income tax, and federal taxes
on dividend income and capital gains).
Equation
2: Wages
| ln
(wage rate) |
= |
+0.0190
ln(government transfers per nonworking adult)
(0.01) |
| |
|
+0.0094
Nonwage labor costs as a percent of the wage
(0.03) |
| |
|
+0.0021
Federal tax rate on labor income
(0.00) |
| |
|
+0.0175
Effective sales tax rate
(0.00) |
| |
|
-0.0021
Cost of capital
(0.50) |
| |
|
+0.3550
ln(working age population)
(0.00) |
| |
|
+
constant terms for each of eight sectors of the economy |
| |
|
+terms
with ln(non-Texas wage rate), for each of eight sectors of
the economy. |
This
equation also gives sensible results, but needs to be interpreted
with care. It shows that when government transfers are higher, the
wage rate will also be higher. A reasonable explanation is that
the safety net gives workers an alternative to working at very low
wages. At first sight it might seem odd that a higher sales tax,
nonwage labor costs, or federal tax rate would raise the
wage rate. Yet the logic is sound; when these taxes are higher,
workers will require higher pre-tax wages in order for it to be
worth their while to work. In each case, however, the after-tax
wage would be lower than before.
Equation
3: Capital Stock
| ln
(stock of capital) |
= |
-0.0389
ln(government transfers per nonworking adult)
(0.00) |
| |
|
-0.0133
Nonwage labor costs as a percent of the wage
(0.08) |
| |
|
+0.0018
Federal tax rate on labor income
(0.18) |
| |
|
+0.0088
Effective sales tax rate
(0.18) |
| |
|
-0.0286
Cost of capital
(0.00) |
| |
|
+1.2760
ln(working age population)
(0.00) |
| |
|
+
constant terms for each of eight sectors of the economy |
| |
|
+
term with index of U.S. output. |
This
equation has no surprises. If capital is more expensive, less will
be used. If sales and property taxes are higher, business will refrain
from investing in the state.
How
does the model simulate the effects of tax changes?
The
simulations are done for 1999 and proceed in two steps. First we
project what the value of the most import variables - employment,
wages, capital stock, and the like - will be in 1999. Then we use
the model to measure what happens to these variables when a change
is made to the tax system. The 1999 baseline values for the most
important variables are summarized below. Further details are provided
in the main report.
| Baseline
Values of Variables Required for Simulations |
| Variable |
Units |
1996 |
1997 |
1998 |
1999 |
Assumptions |
| Labor |
mill.
jobs |
8.725 |
9.092 |
9.401 |
9.617 |
Same
as growth in nonfarm employment |
| Capital |
$
billion |
656 |
707 |
758 |
811 |
Constant
capital/payroll ratio |
| Wage
rate |
$/job/yr |
27704 |
28688 |
29714 |
31109 |
Same
as growth in personal income net of labor growth |
| Working
age population |
thousands |
12274 |
12519 |
12745 |
12974 |
Same
as growth of resident population |
| Welfare
payment per nonworking adult |
$ |
1799 |
1833 |
1869 |
1904 |
Extrapolate
using 1993-96 nominal growth rate |
| Nonwage
labor
costs
as % of wage
|
% |
0.52 |
0.52 |
0.52 |
0.52 |
Extrapolate
using 1992-1996 nominal growth rate |
| Federal
tax on
labor
income
|
% |
18.4 |
18.4 |
18.4 |
18.4 |
Constant
over time |
| Tax
on industrial & commercial prop. |
% |
1.38 |
1.38 |
1.38 |
1.38 |
Constant
over time |
| Sales
tax rate |
% |
6.89 |
6.54 |
6.31 |
6.19 |
Tx
projections; extrapolation for 1999 |
| Cost
of capital |
% |
12.82 |
12.82 |
12.82 |
12.82 |
Constant
over time |
| Source:
From Table 1 in the main report. |
Applying
the model to current tax issues in Texas
A
large number of tax proposals have surfaced in Texas over the past
couple of years. Most of these can usefully be analyzed using Texas-STAMP.
Five of the most important such proposals are summarized below,
followed by a discussion of their economic effects.
| Five
important tax proposals or changes |
| Sales
tax cut |
Governor
Bush recently proposed changes in the sales tax that would,
according to the Governor Bush Committee, amount to a tax cut
of $200 million. Specifically, the change would repeal the current
6.25% sales tax on over-the-counter medicines, diapers and first-aid
items; and it would provide a two-week "back to school"
sales tax holiday for school clothing and footwear. |
| Business
R&D tax credit |
Texas
is one of only five states that does not give businesses a tax
credit for research and development (R&D) or for capital
investments. Intel Corporation leads a coalition that is lobbying
for an investment tax credit. Telecommunications firms, heavily
concentrated in the North Dallas area, favor an R&D tax
credit. The budget office estimates that a tax credit would
cost about $250 million in lost revenue over a two-year period,
which means an annual cost of $125 million. The R&D credit
has the backing of Governor Bush. |
| Raise
the homestead exemption by $10,000 |
In
early 1997, Governor Bush proposed that the homestead exemption,
then at $5,000 per residence, be raised by $20,000 for the purposes
of computing school maintenance and operation (M&O) property
tax. The intention of the proposal was to reduce the fiscal
pressure on homeowners, who had seen their property tax payments
rise rapidly over the previous decade. The direct effect of
the larger exemption would be to lower tax revenue (to school
districts) by $981 million, making this a sizeable tax cut.
The legislature ultimately raised the exemption by $10,000,
which is the policy change whose effects are simulated in this
study. |
| Reduce
the property tax rate by 13¢ per $100 |
As
part of the 1998 package of property tax relief proposals, Governor
Bush has proposed reducing the school M&O property tax by
13¢ per $100 (i.e. 0.13 percentage points), from its average
of about $1.24 per $100 of valuation. The change would apply
to property tax levied on residential as well as commercial
and industrial property. The direct revenue effect would be
to lower takings by $1.11 billion. |
| Exempt
certain businesses from the franchise tax |
The
business franchise tax is levied on corporate income and/or
capital, depending on which method yields the most. The tax
is particularly onerous on small businesses, which sometimes
incur hundreds of dollars in expenses in order to calculate
tax bills that may be as low as $100. In order to reduce the
burden on small firms, Governor Bush has proposed exempting
from the franchise tax all firms with a turnover under $100,000
annually. The direct effect would be to lower revenues by $28.6
million per year. |
The
economic effects of these tax changes are summarized in the next
table. Some additional comments on the findings are in order.
The
cut in the sales tax would lead to 10,953 more jobs. While
this is a large number, it represents an increase of just 0.11%
in the number of jobs in Texas. The "static" effect of
this tax change would be to reduce state revenue by $200 million.
However this would be offset to some degree by additional revenues
that would result from the fact that more people would be working,
and there would be more capital to subject to property tax. The
net effect would be to reduce tax revenue (at the state and local
levels together) by just $182 million.
Summary
Effects of Tax Changes
| A
Cut
sales tax by $200m
|
|
B
Business
R&D tax credit of $125m
|
| |
|
Mean
response |
95%
Range |
| |
|
% |
|
% |
|
|
| Economic
effects |
|
|
|
|
|
|
| Number
of jobs |
+10,953 |
0.23 |
+136 |
0.001 |
92-180 |
| Capital
stock ($ million) |
+900 |
0.22 |
+11 |
0.001 |
7-14 |
| Wage
rate ($/job/year) |
-31 |
-0.20 |
0 |
0 |
|
| Working
age population |
+11,288 |
0.17 |
0 |
0 |
|
| Payroll
($ million) |
+44 |
0.03 |
+4 |
0.001 |
3-6 |
| Revenue
effects ($ million): |
|
|
|
|
|
|
| "Static" |
|
|
|
|
|
|
| Sales
tax |
-200 |
|
|
|
|
|
| Franchise
tax |
|
|
-125 |
|
-125 |
| Property
tax (business) |
|
|
|
|
|
|
| Property
tax (residential) |
|
|
|
|
|
|
| "Dynamic” |
|
|
|
|
|
|
| Sales
tax |
+3 |
|
+.2 |
|
0.16-0.30 |
| Franchise
tax |
+2 |
|
|
|
|
|
| Property
tax (business) |
+13 |
|
+.2 |
|
0.10-0.20 |
| Therefore |
|
|
|
|
|
|
| Net
effect on tax revenue |
-182 |
-1.78 |
-125 |
-0.61 |
-124.5
to -124.7 |
| Memo
items: |
|
|
|
|
|
|
| Total
state tax revenue |
20,564 |
|
|
|
|
|
| Revenue
from sales tax |
12,248 |
|
|
|
|
|
| |
C
Cut
residential property tax via extra $10,000 exemption
|
D
Cut
property tax via rate reduction of
13¢/$100
|
E
Exempt
businesses with gross receipts of $100,000 or less from franchise
tax
|
| |
|
% |
|
% |
|
% |
| Economic
effects |
|
|
|
|
|
|
| Number
of jobs |
+25,449 |
0.265 |
+39,710 |
+0.41 |
+31 |
0.0003 |
| Capital
stock ($ million) |
+4,220 |
0.520 |
+5,059 |
+0.62 |
+2 |
0.0003 |
| Wage
rate ($/job/year) |
+45 |
0.145 |
-5 |
-0.02 |
0 |
0.0000 |
| Working
age population |
+52,990 |
0.408 |
+60,868 |
+0.47 |
0 |
0.0000 |
| Payroll
($ million) |
+1,226 |
0.410 |
+1,186 |
+0.40 |
+1 |
0.0003 |
| Revenue
effects ($ million): |
|
|
|
|
|
|
| "Static" |
|
|
|
|
|
|
| Sales
tax |
|
|
|
|
|
|
| Franchise
tax |
|
|
|
|
-28.6 |
|
| Property
tax (business) |
|
|
-547 |
|
|
|
| Property
tax (residential) |
-491 |
|
-563 |
|
|
|
| "Dynamic” |
|
|
|
|
|
|
| Sales
tax |
+67 |
|
+66 |
|
+.05 |
|
| Franchise
tax |
+10 |
|
+12 |
|
+.01 |
|
| Property
tax (business) |
+58 |
|
+70 |
|
+.03 |
|
| Therefore |
|
|
|
|
|
|
| Net
effect on tax revenue |
-355 |
-1.726 |
-962 |
-4.68 |
-28.5 |
-0.14 |
| Memo
items: |
|
|
|
|
|
|
| Total
state tax revenue |
|
|
|
|
|
|
| Revenue
from sales tax |
|
|
|
|
|
|
Source:
From simulations using the Texas-STAMP model.
Rounding
may occur in the numbers.
The
business tax credit would have a surprisingly small economic
effect - creating no more than about 180 jobs overall - but there
is a clear and logical explanation for this. First, the tax cut
is relatively small; it represents just 0.6% of state revenue, and
would cut the cost of capital to Texas firms by a mere 0.05%. Second,
the tax credit leaves corporations with more profit after tax, but
over a third of this windfall goes immediately to the federal government,
as the higher profit triggers higher payments of corporation income
tax. Third, as the fast-diminishing windfall is paid to owners as
dividends (or accrues as capital gains), it is subject to federal
taxes on personal income, which will again reduce the amount left
for owners. The net effect is to lower the after-tax rental cost
of capital from 14.21% to 14.20%. It is not surprising that such
a small change in the cost of capital will have almost negligible
economic effects.
It
is sometimes argued that if the business tax credit takes the form
of an R&D tax credit, this would ultimately have additional
"downstream effects." This reasoning is based on the idea
that when firms locate their R&D in Texas, then manufacturing
and other facilities will tend to follow. It is not at all evident
that R&D leads and manufacturing follows; an equally plausible
case can be made that manufacturing facilities lead and R&D
follows, in which case the argument for providing tax credits for
R&D (rather than for other forms of investment) is greatly weakened.
The Texas-STAMP model treats all tax credits equally, and is not
designed to measure the possible differential effects of different
forms of tax credits.
On
the other hand, these estimates are based on the assumption that
the tax credit for R&D done in Texas is significantly offset
by higher tax payments to the federal government. However, firms
undertaking new R&D in Texas would benefit not only from the
Texas tax credit, but also from the federal R&D incentive. The
net effect is to lower the cost of R&D quite substantially,
leading to new R&D expenditures for which we do not attempt
to account. Thus our estimates of the effects of the Texas R&D
should be thought of as lower bounds.
The
effect of the $10,000 extra property tax exemption, according
to the Texas-STAMP model, is to increase the number of jobs by 25,449
(+0.265%) and the capital stock by $4.2 billion (+0.520%). The increased
exemption will add $1.2 billion to state payrolls and induce almost
53,000 working-age adults to enter the state.
The
Texas-STAMP model shows that a 13¢ per $100 cut in the property
tax rate would have a substantial economic effect: employment
would rise by 39,710 and the capital stock by $5.1 billion. While
the direct ("static") loss of tax revenue would be $1,110
million, the additional economic activity triggered by this change
would lead to $148 million in extra tax revenue (the "dynamic"
effect), for a net revenue loss of $962 million.
Exempting
businesses with gross receipts not exceeding $100,000 from the franchise
tax would have almost no effect on the economy: The effects
on jobs and capital spending, and therefore also tax revenue, would
be negligible. However the main purpose of this tax cut is to reduce
the burden of paperwork for small firms, an effect that the Texas-STAMP
model is not designed to measure.
These
simulation exercises illustrate the manner in which the Texas-STAMP
model can shed light on the economic effects of changes in the system
of state taxes in Texas. As with any modeling, the results should
be taken as indicative of the orders of magnitude involved, since
it is not possible to predict the effects of tax changes with great
precision.
|