4. Experimento de Viabilidad
4.8. Análisis de resultados
An alternative to an analyst coming up with new state-specific estimates for a program is to rely on the consensus of the national literature. Why rely on hard-to-come-by estimates using state-specific data when there is a national literature with some good estimates of how firms respond to costs?
There is a large research literature on how business location deci- sions respond to taxes. This research literature suggests that for a 20 percent reduction in overall state and local business taxes, on aver- age we would expect a 10 percent increase in business activity.111
This translates into such a tax reduction increasing the probability of a business location decision from 90 percent to the observed 100 percent.
State and local business taxes have typically averaged about 5 percent of the value of a business’s productive activity, what econo- mists call the business’s “value-added.” Therefore, if we assume that what matters to location decision is costs, any 1 percent reduction in costs as a percentage of business “value-added”—which would be equivalent to reducing a business’s state and local taxes by 20 per- cent—would be expected to increase the probability of a favorable business location decision by 10 percentage points, from 90 percent to 100 percent.
Applying National Estimates to Tax Incentives
How might we use these estimated effects of business tax reduc- tions to estimate the effects of tax incentives? We first assume that incentives are simply treated as a cost reduction, as if they are equiv- alent to a reduction in taxes. We then must somehow compare the value of incentives versus taxes, which differ in their timing. The research literature estimates the effects of business tax changes that are ongoing. As described in Chapter 2, incentives are front-loaded, particularly in the first year and the first 10 years.
Some research exists on how business executives discount the future in making investment decisions. Such investment decisions would include business location and expansion decisions. This research literature suggests that business executives heavily discount the future in making investment decisions—that is, their decisions focus mostly on the short term. In making investment decisions, business executives report using an annual “real” discount rate of 12 percent.112 Whatever dollar cost reduction is delivered one year from
now is worth 12 percent less in today’s dollars, even if there were no inflation between now and one year from now. (If there were inflation of, say, 2 percent, then the discount rate would be 14 percent.) This
12 percent compounds over time. As a result, even without inflation, a dollar 10 years from now is worth only 32 cents today.
Why such large discounting? The person at the firm who is mak- ing the location decision probably won’t be working at the firm 10 years from now. Stock prices are heavily determined by short-term profits. An executive concerned with the value of his stock options might be inclined to focus on the short term.
Based on this information, here is one way for state analysts to analyze tax incentives. Take the tax incentive package the state pro- vides to a firm—some schedule of tax incentives or grants by year of the project. Use that schedule to calculate the discounted present value of that incentive package, using a 12 percent discount rate. Also calculate the discounted present value of the firm’s value of produc- tion, its value-added, at a 12 percent discount rate.113 Divide one by
the other to get what average reduction in costs is brought about by the incentives. To get the “but for” for the incentive for that project, multiply that percentage reduction in costs by 10, based on the busi- ness tax literature.114 Box 5.7 provides two more specific examples of
how this methodology can be implemented.115
Some uncertainty exists in these calculations. Perhaps state and local business taxes have lesser or greater effects than the average effects from the research literature. Perhaps businesses respond less or more to cost reductions from incentives compared to cost reduc- tions from business taxes.116 Perhaps this particular firm does not dis-
count the future using a 12 percent discount rate. If the analyst wants to admit uncertainty, the estimated effects on job creation could be cut in half if one is pessimistic, or inflated by one-half if one is optimistic. This methodology has been applied to evaluate a Michigan job creation grant program. This evaluation found that the program had a high benefit-cost ratio under most plausible assumptions (see Box 5.8).
Box 5.7 Two Hypothetical Examples of Using the Tax Research to Estimate Incentive Effects
First, using job figures for the firm receiving incentives, we make assumptions about whether these jobs persist. These examples assume that the original jobs do not shrink or grow over time.
Second, we use data from the U.S. Bureau of Economic Analy- sis on value-added per worker in the industry, and make assumptions about value-added growth over time. For these examples, I use the average value-added per full-time-equivalent (FTE) worker for “trad- able” industries: $194,000/FTE. For simplicity, I assume no growth.
Third, we calculate the present value of value-added over time. We assume the firm uses an annual discount rate of 12 percent. At that discount rate, the present value of value-added is $1.8 million per FTE worker.
Fourth, we calculate the present value of the incentives. I consider two incentives:
1) An up-front job creation tax credit of $20,000 per FTE. 2) A 50 percent property tax abatement for 10 years. At average
property taxes, this abatement is worth about $2,300 per FTE per year.
The present value of the job creation tax credit is $20,000. Using a 12 percent discount rate, the present value of the abatements is $14,500. Fifth, we calculate the present value of incentives as a percent of the present value of value-added. The job creation tax credit is about 1.1 percent of the firm’s value-added. The property tax abatement is about 0.8 percent of the firm’s value-added.
Sixth, we assume how sensitive business location decisions are to lower costs. The research consensus is that 1 percent lower costs increases the probability of tipping that location decision by 10 times as much. Based on these assumptions, the job creation tax credit will tip 11 percent of location decisions. The 10-year property tax abate- ment will tip 8 percent of location decisions.
NOTE: These calculations use the 31 tradable industries in Bartik (2017a). BEA data is used to calculate value-added/FTE. The database in Bartik (2017a) is used to calculate average business property taxes.
Box 5.8 Evaluation of the Michigan Business Development Program
Begun in 2012, the Michigan Business Development Program (MBDP) has provided $180 million in grants to 239 projects. The grants are an up-front subsidy for job creation averaging $7,500 per job. About 39 percent of incentive dollars have gone to auto firms, and 34 percent to other manufacturing firms.
This evaluation applied a version of the incentive simulation model of Chapter 4 to MBDP. This included estimating a plausible “but for” and multiplier, and estimating economic effects under differ- ent methods for financing MBDP.
The evaluation found a benefit-cost ratio for MBDP exceeding 4-to-1 if MBDP is financed by higher taxes. If MBDP is financed by reducing public school spending, MBDP’s net benefits are negative.
MBDP’s benefit-cost ratio is higher than typical incentives. This higher benefit-cost ratio occurs for three reasons:
1) MBDP’s incentives are modest, which increases their effec- tiveness per dollar.
2) MBDP’s incentives are up front, also increasing effectiveness. 3) MBDP is targeted at high-multiplier industries.
The evaluation also considered whether MBDP would pass a benefit-cost test if the “but for” effect were significantly less than the research consensus. Because of high multipliers and the incentive design, MBDP would have net benefits even with a smaller “but for.”
NOTE: See Bartik et al. (2019) for more details on the example in the text box.
Customized Services
We can also apply national estimates to customized business ser- vices. As mentioned above, we can plug customized services into our simulation model, assuming that their job creation effects are 10 times that of tax incentives. This yields a cost per job created of $46,000.
If this estimate is valid, estimating program effects is simple. Determine how much in customized services the program is provid- ing each year, and divide by $46,000 to get a plausible estimate of job creation in incented firms.
This estimate relies on assuming that the state program being analyzed is of comparable quality to the programs being studied in the national estimates. The national studies were looking at exemplary programs. Perhaps this particular state’s program is of lower quality. To be conservative, the analyst might want to do some downward adjustment in jobs created. How much? That’s hard to say without more knowledge about the quality of the state program. There’s only so much one can do without having some specific data or other infor- mation on the quality of a state program.