August 17, 2017

Less Optimistic Assumptions Delay the Breakeven Date for Proposed Subsidies and Make Reaching that Point Less Likely

Foxconn has said that the new plant it wants to build in southeast Wisconsin could employ between 3,000 and 13,000 workers. Whether the number of employees is ultimately at the top of that range or far lower will dictate how much the proposed subsidies would cost Wisconsin taxpayers per job and how long it would take our state to break even on the new tax credits.

According to a Legislative Fiscal Bureau (LFB) summary of a recent Department of Administration analysis, it is likely to take until at least 2043 for the cost of the new state tax credits for Foxconn to be fully offset by the income and sales tax revenue generated directly or indirectly from the new enterprise. That breakeven period is based on what has often been described as the “best case scenario” – with 13,000 people employed at the new plant by Foxconn and 22,000 new jobs elsewhere resulting indirectly from the new economic activity. It also assumes that 100 percent of the new tax revenue is captured by Wisconsin, despite the fact that many of the workers and suppliers may be in Illinois.

Using the same methodology but somewhat less optimistic assumptions about how many jobs are created and how much of the new tax revenue accrues to Wisconsin, this report describes two alternative scenarios in which the breakeven point would not come until 2050 or 2058. We also examine how much Wisconsin could be in the hole if the Foxconn plant were to close well before we reach the breakeven date.

The Foxconn Job Numbers

According to Foxconn, the proposed plant will employ about 3,000 workers within a few years and their workforce could grow to as much 13,000 employees. However, no guarantees of that have been provided to policymakers who are being asked to approve billions of dollars in state and local funds to help make the project happen. The “memorandum of understanding” (MOU) between the state and Foxconn merely says that Foxconn will employ “up to” 13,000 workers.

Based on various assumptions provided by Foxconn, the accounting firm it hired (EY, formerly known as Ernst and Young) estimated that building the new plant would require 10,200 construction workers, and EY projected that the economic activity would generate another 6,000 indirect or induced jobs during the construction period. Even more importantly, they estimate that the ongoing operation of the plant would yield 22,000 “indirect and induced jobs.”

Many commentators have noted that employment and payroll projections are often exaggerated and that Foxconn does not have a good track record of keeping its promises about job creation and job permanence. As the Fiscal Bureau report pointed out, the EY study contains a noteworthy disclaimer about the assumptions it was working from: “EY does not express an opinion… as to whether the underlying assumptions provide a reasonable basis for the (EY) analysis.”

The Structure of the Proposed Subsidies

The proposed legislation would authorize Foxconn to receive up to $2.85 billion in new tax credits over a 15-year period. Although these subsidies are referred to as tax credits, they are expected to take the form of very large checks written annually by the Department of Revenue to Foxconn because existing corporate tax breaks make it highly unlikely that the company will owe any corporate income tax against which these refundable credits would be claimed.

For Foxconn to receive the maximum amount of new tax credits, it would need to spend more than $8.8 billion for payroll and at least $9 billion for capital investments in its new facilities. However, as the proposal is currently structured, it does not preclude Foxconn from using the new capital expenditure tax credits to execute a business plan that stresses state-of-the-art robotics, with a much smaller investment in payroll. Under such a plan, Foxconn would receive a smaller amount of total credits, but the state would pick up a much larger portion of the cost for each new worker. In addition, there would be less new tax revenue generated, and the breakeven period would be pushed further into the future.

Because it is unclear how many jobs will be created, and since the legislation allows the emphasis to be on robots rather than employees, it is important for state policymakers to think through the risks and rewards of the legislation under a number of different scenarios. Legislators and interested state citizens who will foot the bill should not pin all of their hopes on a single scenario that could easily prove to be overly optimistic.

Alternative Scenarios and Key Assumptions

This report examines the breakeven period under three different job creation scenarios, using essentially the same methodology employed by the DOA in the analysis summarized by the Fiscal Bureau (as best we were able to reproduce the calculations they made). These scenarios are not forecasts of what we expect to happen, and we also caution people not to construe the LFB’s summary of the DOA analysis as a forecast. Instead, these are illustrations of how the payback period for the state’s investments is dependent on key variables that are beyond the control of state policymakers.

Our first scenario is exactly what DOA assumed, sometimes described as the best case scenario. Before explaining the ways in which our second and third scenarios are different, we should make it clear that most of our assumptions are the same. We follow their lead in assuming the following:

Foxconn will spend at least $9 billion for capital investments and will receive $1.35 billion of investment tax credits (in addition to whatever job tax credits it receives).
Over the next several years there would be 10,200 construction-related jobs (including employment with equipment suppliers), and 6,000 more indirect or induced jobs resulting from the construction spending.
The average annual wage for the Foxconn employees will be $53,875.
The average annual compensation for the construction-related employment will be $59,600 (including benefits).
The construction-related spending over a 4-year period will generate $187 million in increased state taxes (almost entirely from income and sales taxes).
On an ongoing basis there will be a significant number of additional jobs created by the economic activity, and these indirect or induced jobs will have an average annual wage of about $51,000.
For every 1,000 indirect or induced jobs, state tax revenue will increase by about $3.2 million.

The areas where our assumptions diverge are the following:

The “best-case scenario” (i.e., the DOA assumptions)

Foxconn employs 13,000 full-time workers.
The new economic activity yields 22,000 indirect or induced jobs with other employers.
All of the increase in tax revenue is captured by Wisconsin.
The state’s costs include $10 million set aside for local grants, but no increase in spending for training.

Alternative 1

There are 6,500 Foxconn employees (50 percent of what DOA assumed).
There are 16,500 other new jobs (75 percent of the jobs and payroll that DOA assumed).
95% of the growth in state taxes is captured in Wisconsin.
The state spending also includes $20 million that the amended bill commits to training.

Alternative 2

Foxconn employment is at the low end of the range they have discussed, 3,000 full-time workers, because they invest more heavily in robotics.
There are 11,000 other new jobs (50 percent of the jobs and payroll that DOA assumed).
90 percent of the growth in state taxes is captured in Wisconsin.
The state spending includes $20 million that the amended bill commits to training.

The following table illustrates a couple of the assumptions that our shared across all three of the scenarios we have examined, but it focuses primarily on the where the assumptions diverge, and it also summarizes our findings about the implications of different assumptions for the estimate breakeven dates.

Table 1: Less optimistic assumptions postpone breakeven date

	Best Case Scenario	Alternative 1	Alternative 2
Foxconn jobs	13,000	6,500	3,000
Construction period jobs	16,200	16,200	16,200
Other “indirect” jobs	22,000	16,500	11,000
Total ongoing jobs	35,000	23,000	14,000
Cost of tax credits	$2.84 billion	$2.36 billion	$1.73 billion
Training funds & local grants	$10 million	$30 million	$30 million
New taxes, construction period	$187 million	$187 million	$187 million
New taxes, ongoing	$115 million/yr.	$75.3 million/yr.	$46.7 million/yr.
New tax growth captured by WI	100%	95%	90%
Breakeven year	2043	2050	2058

Estimating the “Indirect” Employment Effects

It is easy to take different assumptions about the number of Foxconn employees and compute what each of those scenarios would mean for the amount of new tax credits and the cost of those tax credits per job. However, calculating the effect on state tax collections and how long it will take for growth in state revenue to offset the cost of the tax credits and other subsidies is far more difficult because a lot of other variables and assumptions come into play.

One of the important questions is the matter of making an estimate of the relationship between Foxconn’s hiring and payroll and the employment gains that Foxconn’s spending will generate elsewhere. The consulting company’s analysis for Foxconn estimates a multiplier effect of 2.7. What that means is that the creation of 13,000 jobs at Foxconn would generate an increase of 22,000 jobs elsewhere, for a total of 35,000 jobs.

With that in mind, the easiest way to consider scenarios with fewer workers at Foxconn would be simply to apply the same multiplier in each case. However, there are two major components of the multiplier effect. One is the employment at businesses that supply components or services to Foxconn. The other component is the positive impact on the local economy of the spending by Foxconn employees. If Foxconn builds a factory with state-of-the-art automation and fewer employees, that will significantly reduce the positive effects of having an influx of employee earnings flowing through the economy, but it will not necessarily cause a large decrease in Foxconn’s contracts with other businesses. (But keep in mind that those businesses will not all be in Wisconsin.)

With those considerations in mind, we decided to assume that a 50 percent reduction in the number of Foxconn workers would result in a 25 percent reduction in the employment gains resulting from the new economic activity. And for alternative #2 we assumed that if there are just 3,000 Foxconn employees there would be a 50 percent reduction in the indirect employment gains.

We do not claim to possess great economic expertise underlying our assumptions. This report is an exercise in risk analysis rather than an effort in economic forecasting. In other words, it is a mathematical exercise to illustrate some of the ways in which projections based on one set of assumptions can be far off the mark if those assumptions are incorrect.

To What Extent Will Illinois Share in the Tax Gains?

Another one of the significant challenges for anyone attempting to calculate a breakeven period for a set of business subsidies is the question of how much of the increased economic activity will be in our state and how much of the increase in tax revenue will be reaped outside our state. That’s especially relevant in this instance because the new plant will be very close to the Illinois border, and northern Illinois has a significantly higher unemployment rate than southern Wisconsin.

The DOA analysis assumed that 100 percent of the new state tax revenue would go into the Wisconsin treasury, which seems unrealistic. Many of the workers at the Foxconn plan will commute from Illinois, and some of the new income they are earning will generate sales tax revenue south of the border. In addition, some of the new “indirect” employment that the DOA analysis is counting on will almost certainly be in businesses located outside Wisconsin. As a result, some of the indirect employment with Foxconn suppliers will boost Illinois income and sales tax collections rather than Wisconsin’s tax coffers. With that in mind, our alternative scenarios use slightly less optimistic assumptions about how much of the new tax revenue will end up in Wisconsin.

Under “alternative 1” we calculate that although the Foxconn subsidies would be lower, it would take until 2050 for increased tax revenue to fully offset the costs. And under “alternative 2” – which decreases the number of Foxconn jobs and assumes that 10 percent of the growth in tax revenue would end up in the coffers of Illinois or other states – the breakeven point would be in 2058. If the plant relies heavily on robots and is at the low end of the range of employees suggested by Foxconn, less of the indirect employment will result from workers’ spending in the community, and a larger percentage will be the jobs with suppliers who could be outside of Wisconsin. Under that scenario, it wouldn’t be surprising if at least 10 percent of the increased income and sales tax revenue is reaped outside Wisconsin.

The Limitations of These Sorts of Calculations

We have attempted as best we are able to perform the same calculations used by DOA in the analysis summarized by the Fiscal Bureau. Without being privy to all of their assumptions about the mix of indirect jobs that stem from the new plant, our calculations could be slightly different than theirs (if they were using our assumptions about Foxconn jobs and the tax revenue reaped outside Wisconsin). However, we are confident that our estimates of the breakeven period are very close to what DOA or the LFB would arrive at if they analyzed our alternative scenarios.

A more basic problem with the methodology that we borrowed from DOA is that it does not consider other costs, such as the cost of diverting state and local resources from other uses. A strong case can be made that the breakeven point under any of the three scenarios we examined would actually be even further in the future if it were conducted by economist who assigned a cost to tying up public resources for this project instead of having those dollars invested in other ways.

Other costs are also omitted. In contrast to the DOA calculations, we did include in our analysis the $20 million that the proposed legislation would set aside for training of Foxconn workers. However, we did not include in our calculations any portion of the substantial costs of improvements to the Interstate corridor. Nor did we attempt to put assign a cost to the potential environmental problems that could result from this type of factory, especially since the state plans to short circuit some of the normal environmental protections.

The Risk of a Premature End to Foxconn’s Wisconsin Operations

Considering how quickly television and other flat screen technology has been changing, there is a significant risk that the proposed Foxconn plant will not be operating long enough to come close to offsetting the state’s massive investments. Assuming the new plant does not become obsolete even before 2034, that year is when the risk of Foxconn pulling up stakes and going elsewhere is probably greatest because it is when the state would stop writing the company huge annual checks for 17 percent of payroll costs. When a company’s business plan relies on such large subsidies, how confident can we be that they will remain in Wisconsin when those subsidies are gone?

Table 2: Estimate of the unrecovered subsidies if the Foxconn plant closes

	Best Case Scenario	Alternative 1	Alternative 2
Foxconn jobs	13,000	6,500	3,000
Unrecovered subsidies:
Foxconn plant closes in 2034	$922 million	$1.10 billion	$949 million
Foxconn plant closes in 2042	$1.8 million	$530 million	$621 million

One of the surprising findings of our analysis is that the net loss for Wisconsin taxpayers from Foxconn exiting Wisconsin in 2034 is in the same ballpark regardless of which of the three scenarios one assumes. Table 2 shows our estimates of the net cost to the state under each of the three scenarios if the Foxconn plant is no longer operating in 2034 or if it ceases operations 25 years from now (in 2042). The difference between the three scenarios we analyzed becomes much clearer in later years, after the state stops paying job credits that are directly proportional to the size of its payroll.

Conclusion

Our calculations help drive home the point that it may be wishful thinking to believe that the huge subsidies Wisconsin lawmakers have proposed granting to Foxconn would be offset by increased tax revenue by 2043. And our analysis reveals that if the plant closes much before then, Wisconsin won’t come anywhere close to recovering what it spends for the annual checks that will be sent each year to Foxconn.

Although we think our calculation of the effects of alternative assumptions can help policymakers and the public balance the potential risks and rewards of the proposed deal, our use of the DOA methodology has a significant shortcoming. We are underestimating how long it would take to reach the breakeven point because these calculations do not account for some of the other costs, such as the significant costs of not using the state funding in other ways that could yield significant returns.

Alternative investments of the $2.85 billion for new tax credits could also have far less risk for taxpayers than making a massive bet on a single enterprise that could go out of business or move elsewhere long before increased state tax revenue offsets the huge costs incurred over the first 15 years of operation.

Jon Peacock

The Significant Risk of Not Breaking Even on Foxconn Subsidies