Arizona Fiscal Issues
Response to: Beacon Hill Institute's Tax Analysis Modeling Program: A Response to Charney
By Alberta H. Charney, Ph.D.
May 6, 2010
Introduction
Point 1
Point 2
Point 3
Point 4
Point 5
Point 6
Point 7
Final, smaller points
End Notes
Introduction
I computed job impacts for the proposed 1-cent increase in the sales tax, in which I compared the job loss associated with the sales tax increase with the job loss associated with the already passed contingency cuts in the state budget if the measure does not pass. My results showed that increasing the sales tax will reduce jobs by 7,400 jobs but that making the comparable amount of cuts in the state budget will reduce jobs by 20,500 jobs. Thus, increasing the sales tax to prevent the cut in the state budget was found to retain approximately 13,500 jobs in the state's economy. When asked why my results were so different from the Goldwater Institute's press release, whose impact was computed by Beacon Hill Institute's STAMP model, I did an analysis entitled "Comparison of UA, REMI, and STAMP Simulations of Tax/Spending Increases.' Links to the original study and the comparison study are below. The Beacon Hill issued a response to my analysis, which can be linked to through the Goldwater Institute's website. I refer to their response document as BHIR. These are my responses to their comments.
Point 11
When are the Goldwater Institute (GI) and/or the Beacon Hill Institute (BHI) going to take responsibility for releasing a set of numbers without any explanation of what numbers were run and why their results are so different from both IMPLAN and REMI. They are the outliers. It is not responsible to post a press release with a link to a one-half page undocumented table.
Even in their response, they do not take the time to discuss exactly what numbers were entered into their model and which economic activities are directly affected by the tax in their simulation. In my "Comparison...," I expressed concern that their simulation had not correctly taken into account all the exemptions in Arizona's taxable sales, but I had implicitly given them the benefit of the doubt that they had collected Arizona taxable sales data and introduced proportional changes in the corresponding categories in their STAMP model for Arizona. But apparently this was wrong. In their response statement, they said:
"A STAMP model accounts for the relevant state's sales exemptions in the different sectors of the state it models. That is why we modeled proportional sales tax increases in five different states to calculate the average percentage change in each economic variable that we then applied to Arizona BHIR, p.7."'Â
The first sentence is very clear, but I have no idea what measures they are averaging from other states nor how those averages are used in STAMP. Are the Arizona results just the average of simulations of five other states? Did they average taxable sales from five other states and input those into their model for Arizona? Why would they do this, given that taxable sales by category are available for Arizona?  I don't know what this argument means. Further, this statement does nothing to inform us as to the exact magnitude of taxable sales changes they introduced into their model; if anything, this comment makes that question even more important. It leads one to wonder whether they have a single generic STAMP model that they apply to all states so they have to try to figure out how to adjust the generic data to approximately represent Arizona.
Further, they have never explained why their one-half page results table was titled "A $1 billion increase in transaction privilege use and severance tax revenue (Scenario 1)." Because I questioned whether or not BHI incorrectly increased some non-taxable activities in their simulation, their they claim in their response that "In our simulation we assume that the sales tax falls only final goods consumed in the state (BHIR, p.7)." So why did they simulate an increase in severance taxes?
Point 2
BHIR implies that since our findings indicates a positive job impact of raising taxes to prevent major cuts in state services, then why doesn't Arizona just keep increasing taxes indefinitely thereby increasing net jobs. They should know that models can only be used for modest changes from existing economic conditions and that results from modest changes cannot be used to predict what would happen with large, never before seen, changes in policies.
Further, I want to say clearly and without reservation that no one, particularly me, is arguing that Arizona should raise taxes indefinitely because government spending creates more net jobs. Nor would I argue that Arizona should allocate more on spending categories that would result in inflows from the federal government, just for the sake of having more money flow into the state. What I have argued for a long time, and will continue to argue, is that Arizonans need to decide what level of public services they want and create a tax structure that raises the necessary funds to pay for them that is as efficient and fair as possible.
If they are going to make that accusation, however, isn't the reverse question applicable? Since the BHI STAMP model always results in negative economic consequences of a tax increase then, for the benefit of Arizona's economy, shouldn't all taxes be immediately repealed? If tax cuts always create growth, then Arizona should be flying high. Arizona has had 42 cuts to the individual income tax, sales tax and corporate income taxes since the early 1990s. Does anyone really believe that the march to the bottom in terms of taxes and government services has really increased state growth? Have all these tax cuts immunized the state against downturns? (More about this taxation and growth will be discussed below.)
Arizona is a state that has benefited enormously from government investments throughout its history, yet there are those who would deny that government activities are important to the well-being and continued development of the state. A major portion of the land we currently call Arizona was purchased from Mexico by the federal government. And where would Arizona be without the Roosevelt Dam, the interstate highway system, the Central Arizona Project, our regional highways, or the universities, community colleges, and the entire educational structure in the state?
Point 3
One of BHI's arguments is that "states compete with each other for residents and business and that state taxes negatively affect state competitiveness." The STAMP model incorporates every conceivable negative consequence of taxes that can be built into a model, regardless of the level of taxes in the state. It is built to compute negative tax effects. In the theoretical literature, relative taxes affect resident and business decisions. Consider a hypothetical state that has the lowest tax rate among the 50 states. Suppose that this hypothetical state raised taxes but still remains the lowest tax state in the country. If STAMP were applied to that state, it would still predict negative impacts of the tax increase, even though that states' relative taxes are still the lowest. There would be a concern about Arizona's competitiveness if it were among the high tax states, but it is not. Arizona ranks very low in taxes overal - typically in the lowest fifth of all states,2 depending on who/how the rankings are done, so competitiveness on the basis of relative taxes is not a major issue.
A more important, but related, point is that while states compete with each other on the basis of taxes, they also compete on the basis of the level/quality of public services. Public services have been found time and time again to exert positive effects in empirical regional growth models but this is completely ignored by the STAMP model. In regional growth models, the positive effects of public services on growth offset the negative effects of taxes to a less or more degree depending on the state and the tax/spend combination. Because both tax variables and public service variables have been included in empirical regional growth models, the negative coefficients on the tax variables have been badly misinterpreted (see the discussion below).
Point 4
The underlying foundation of the BHI analysis is faulty, based on the introduction to their model. The BHI STAMP PA model description of the STAMP model for Pennsylvania report begins by stating:"Changes in tax rates have measurable effects on taxable activities. The weight of evidence shows that state-level tax increases have significant negative effects on state economic activity (p.3)."
The source information to support this statement (it was not a quote) was a 1991 book by Timothy J. Bartik3. The following is Bartik's actual conclusion, based on his analysis of 84 studies of the effects of state and local taxes on business activity:
"The review of previous research suggests that state and local policies can have significant effects on local growth. A state and local business tax reduction of 10 percent, without reducing public services to businesses, probably increases business activity in a state or metropolitan area in the long run by 2.5 percent. Improved state and local public services to business can increase growth. Some evidence suggests that increasing state and local taxes to finance improved business services will have a net positive effect on local economic growth (Bartik, 1991, emphasis added by current author)."
Based on this statement, I would not be able to argue that "the weight of the evidence shows that state-level tax increases have significant negative effects on state economic economy." I ask the reader to read Bartik's quote again and focus on the part that says : "without reducing public services to businesses." This is a critical point in my comments and I will be returning to this again.
Bartik also published an overview of these same studies in 1992.4 The conclusions in that overview are more specific:
"Lowering state and local business taxes can significantly increase business activity in a state, metropolitan area, or suburb, but such a strategy may have large costs. For a state or metropolitan area, the annual costs of a business tax reduction strategy per job created are probably somewhere between two and eleven thousand dollars - that is, the costs range from 'expensive but maybe worth it' to 'much too expensive.' Business tax reductions financed by cutting education or infrastructure spending will probably destroy jobs (Bartik, 1992, p.109, emphasis added by current author)."
In a separate report, "Public Services Positively Impact Economic Growth: A Review of Taxation and Growth Literature," I review this literature and draw the following conclusions:
- The phrases so commonly heard, that "higher taxes hurt growth" and that "lower taxes stimulate growth," are complete misrepresentations and oversimplifications of the taxation and growth literature.
- Both public service and tax measures are included in growth model equations, but only the effect of the tax variable is discussed or interpreted. Impacts of the public service variable are largely ignored.
- The level of public services is shown to have an important positive impact on growth.
- Because in these models growth is negatively affected by taxes and positively affected by the level of public services, then the correct interpretation of the impact of taxes on growth is: "lower taxes may increase growth, but only if public services levels are not cut at the same time."
- Including public service measures in growth models increases the size of the measured negative tax effects.
- Removing public services from growth models renders the tax variable statistically insignificant.
Point 5
I will repeat the main point that I made in my original study and in my STAMP critique --
namely, that governments spend money differently than individuals and the job impact of government spending is much higher than that of the spending of individuals. In particular, BHI makes it very clear that, in STAMP, government doesn't seem to hire directly nor does it spend money in a way that would produce direct jobs in the private sector (via government spending). According to BHIR, STAMP "allows that a large portion of the new money would find its way back to households, which would in turn spend the money and, in that process, ‘create' new jobs in the private sector (BHIR, p.4).' Further, BHIR, p.5 states: "Thus tax money that we allocate to households goes into spending (and job creation) as effectively as it would if we allocated it to government to provide the same services.'Â
The whole point of my analysis is that households don't buy the same services as government and so job creation is, in fact, very different. If all government does is take money from individuals, buy some materials, then give the rest back to (same or different) individuals to spend, then IMPLAN, REMI and STAMP would all result in net negative impacts of a sales tax. Rather, governments provide services either through direct funding and hiring (such as education) or buying services from the private sector (such as health care) that hires people. All these services employ more people per million dollars of expenditures than most other types of household expenditures. I don't think this is a hard concept for most people to understand. Using an extreme example, a one million dollar purchase of wide screen televisions will not create as many jobs in Arizona as spending one million dollars on education because the televisions aren't produced here and only the retail margin is retained in the state.
STAMP lacks a sufficient mechanism to convert government spending to actual jobs. Although BHI recognizes that governments provide "services,' their model doesn't directly allocate workers to provide those government services. In STAMP, much of government is represented as a big pass-through device where tax dollars come from households and money goes back out to households resulting in the same the per dollar spending effects. BHIR can say I have a "jobs fixation' if they like, but most people recognize that when $900 million is cut from state spending, there will be fewer teachers, fewer medical personnel, fewer day care workers, etc. But in STAMP, when "government services' are cut there are no jobs directly connected to the provision of many of those services. BHIR make the argument that "If Medicaid pays $100 for a private-sector doctor to treat a poor person, the job-creation effect of that payment is exactly the same as it would be if the government hired its own doctor to provide the treatment (BHIR,p.6).' That is exactly correct. The problem is that, in STAMP, neither Medicaid nor state government hires the doctor directly; instead, the money is just given back to households. The model doesn't include or recognize that the necessary mechanism to get money back to households is through employment, either in the public sector or in the private sectors affected by government spending.
Point 6
Since STAMP is a long-run model, then it ought to contain characteristics that determine long-term growth rates. STAMP includes public infrastructure in the model, where "The infrastructure capital stock for the current year is the infrastructure for the previous year, less depreciation plus the net spending on transportation by state and local governments (STAMP, p.27)."
Unfortunately, there is complete silence in the PA STAMP document about investment in human capital. They model nothing about the long-term effects of public education expenditures on human capital or on the supply of different labor skills. The accumulation of human capital can be every bit as or more important for growth as the accumulation of physical capital (both public and private). This is a major shortcoming for a model that is all about incentives, investment decisions and long-run production effects. Education is a major portion of state and local government expenditures and yet its role in long-run accumulation of human capital is totally ignored. Thus they ignore the long-run economic disadvantages of an area that has a poorly educated workforce associated with major reductions in education expenditures.
Point 7
Some STAMP elasticities are indeed arbitrarily set. In the "Comparison..." study, I said that some of the import elasticities are set arbitrary in the STAMP model. BHIR responds that elasticities were taken as found from the literature.  As I commented in "Comparison …' earlier, they bear no relationship to the average of the elasticities in the referenced article (I previously indicated I have not reviewed the book). But the main problem is that those referenced elasticities were computed for tradable (across national borders) manufactured products. Manufacturing sectors represent 14 of the 27 non-government sectors in the STAMP model. The rest of the sectors (which represent the bulk of the economy) have no estimated import elasticities in the literature, so BHI arbitrarily set them. They could have omitted import elasticities for the sectors for which they had no estimates, but chose not to.
In STAMP, all the manufacturing import elasticities are set at 1.5. They state "The reason for the uniformity of the elasticities of imports across sectors is that we could not find more detailed estimates in the literature, BHIR, P. 8." I don't understand this because, for the manufacturing categories, they had detailed manufacturing import elasticities from the article BHI cites. They could have tailored them to Arizona, by sector, by computing weighted averages based on the mix of actual Arizona manufacturing (averaging in the zeros for sectors with negative elasticity estimates, as I discussed in "Comparison..."), but they did not.
Final, Smaller Points
The STAMP model, because it is a CGE model, is inherently a long-run model. This raises the question of whether STAMP or any CGE model is the appropriate tool with which to model the short-term economic effects of a temporary tax.
I fully understand the differences between input-output models (IMPLAN) and computable generalized equilibrium models (CGE, including STAMP). Apparently I offended BHI sensibilities by referring to both as impact/policy models. The use of the descriptive word "impact' is to distinguish those models from forecasting models or some combination type of model, such as REMI, that provides impacts over time in a forecasting framework. The sole purpose of a CGE model is to assess the change in the computed equilibrium before and after a shock (in this case a change in taxes), so it computes "impacts.'
STAMP does not define all variables – just one example, ENTRHO, page 34.
In BHI's reference to public workers, they make the side comment "...even given the high pay received by those workers (BHIR, p.4)." In Arizona, state and local government wages and salaries per employee are less than wages and salaries per employee for private sector workers.
For further information and analysis on these important topics please refer to Dr. Charney's previous articles in this area:
- What Will It Cost If Arizona Voters Reject the One-Cent Sales Tax Hike on May 18, 2010? 3/30/10
- Sales Tax Increase vs. Expenditure Cuts: An Economic Impact Study. 3/19/10
- Comparison of UA, REMI, and STAMP Simulations of Tax/Spending Increases. 3/19/10
For more background please refer to EBR's Arizona Fiscal Issues webpage.
End Notes
1. These points are in no particular order and they do not correspond to Beacon Hill's comments, which were not numbered.
2. See the discussion of Arizona's rankings in Tom R. Rex (2009), The Tax Burden In Arizona, a Report from the Office of the University Economist, Center for Competitiveness and Prosperity Research, L. William Seidman Research Institute, W.P. Carey School of Business, Arizona State University.
For further information, please contact us.
