The Future of SALT(?): Using and Improving Formulas for Apportioning Income, Residence and Intangibles

This data raises a simple question: How can the proportion of sales in California across all corporations be so much less than California’s share of U.S. GDP (between 13% and 14%) or population (over 11%)? The stakes are high. California would essentially double its corporate tax revenue if its sales factor represented its share of the economy because this is now the only factor that California uses.

We even have recent evidence of an anemic sales factor for California from the tax records of one large, very profitable, multinational corporation: Microsoft. As part of unrelated litigation, Microsoft revealed that its sales factor for California ranged from 5.9% to 7.3% for the years 2015-18. This is generally right in line with the aggregate estimates provided by the FTB.

These results may indicate that state formulas are defective in a fundamental way that does not implicate their design or taxpayer behavior. Perhaps predominantly market states have chosen to use just the property factor while predominantly industrial states have chosen to use the sales factor? Such choices could result in systematic under apportioning. As a matter of political economy, this seems unlikely.Does it really makes sense to argue that California has an anemic sales factor relative to its population and GDP because Californians systematically consume less than residents of other states relative to their wealth and population?

It seems much more likely that California has adopted the sales factor because it is a wealthy state and its politicians expected it sales factor to reflect it consumption power. Yet it does not. This suggests there is something about the design of the sales factor that is an issue.

Now, it could be just the design is flawed but there are no taxpayers exploiting these flaws, but this also seems unlikely. At least two prominent insiders have explained that tax planners take apportionment into account. In most of the examples provided by these insiders, the “efficient” tax result is achieved through manipulation of entities rather than gaming the formulas themselves, but as early as 2009 one of these practitioners, Barnwell, explained that “the ability to implement powerful apportionment planning with a focus on the sales factor has emerged as the new fulcrum.” It is hard to believe that taxpayers are not using this new fulcrum to contribute to the anomalous results in the data reported above. After all, the evidence of income shifting at the national level is well known, if not wholly settled, and it is hard to see why tax planning should somehow stop at the water’s edge. If Microsoft had simply used the population default provided in California regulations, for instance, its apportionment factor would have been much higher.

On the anecdotal side, another prominent former practitioner, Griswold, lists over 30 reported cases detailing the tax planning maneuvers he describes in one article. And the numbers in these cases are significant: in one case of one large company in one state, the reduction in liability was from about $5 million in self-assessed liability to zero. Another striking aspect of that case was that it emerged from the taxpayer’s own tax department. It is hard to believe that such firms are outliers or that there was only a widespread culture of aggressive tax planning in one firm, especially since making favorable judgment calls in connection with the sales factor likely does not require similarly aggressive choices in many cases.

Thus, available evidence suggests that the problem with the sales factor, at the moment, is primarily an issue of under-reporting. Under-reporting results in the problems with fairness and efficiency already discussed. Striving for improvement is warranted. To be sure, the goal cannot be perfection, nor should it be, given the cost that outcome would impose to federalism values because it would presumably need to be imposed from the center. Nevertheless, the problems with the sales factor are beyond what states should accept.

In addressing these issues, states should resist the calls of faux precision. Taxpayers and their advocates continue to propose using self-generated economic models as a solution, but, again, this generally self-serving approach is inconsistent with the whole formulaic enterprise and should be rejected. The push to use such approaches is another, admittedly rather impressionistic, indicator that there are taxpayers interested in making significant investments to drive down their sales factors and there are tax professionals who would like their business. What, therefore, is to be done?

C. The General Approach

Recently, the United States imposed a minimum tax that relies on accounting income as a backstop for taxable income. Reforms at the OECD level make similar use of accounting concepts. Now, on the one hand, this shift can be criticized as allowing the failure in one domain (tax) to spread to others (accounting). If a country does not believe the tax system is producing an accurate result, then the tax system should be changed. There is surely something to this argument, but there is another justification for the use of accounting information, one that would be true even if we improved the tax system (as we should): because accounting income is enmeshed in a different system, subject to different incentives, it will always be a reasonable backup.

One way, therefore, to improve the functioning of apportionment formulas is to integrate information from other complementary domains—at least as a backstop. If the information is (more) accurate, then we improve accuracy and, if it is collected in another domain anyway, then the increased accuracy is achieved with little cost of administration. We can go a step further. It is inherent to modern tax administration to rely on some form of third-party verification. When that is done, compliance is rather high. When it is not done compliance is much lower. Permitting taxpayers to self report their sales is not so different from allowing them to self report their income. Insisting on third-party verification for the information used in formulas, or, failing that, requiring the use of more granular general information developed by third-parties can be seen as bringing the use of formulas up to the late-twentieth century state of the art.

The OECD has the same idea of using accounting information to calculate income for purposes of what is known as Pillar One, a new tax regime that would apply to many of the largest and most profitable multinational corporations. The OECD also proposes the use of such information or similar information in connection with Pillar One’s allocation rules. Specifically, taxpayers must use an “Enumerated Reliable Indicator,” defined, in part, as an indicator “relied upon by the Covered Group for commercial purposes or to fulfill legal, regulatory, or other related obligations.”

Some further conceptual arguments for the preferred approach are worth noting before diving into details. Let us suppose that there is useful information for apportionment purposes reported as part of federal securities law. There are two related channels by which this could be helpful. First, a business wants to—and must - represent its current business operations accurately, just as it does not want to understate its income and spook its investors. Thus the securities data, because of its different incentives, might be more accurate or at least serve as a kind of check.

Alternatively, one might argue that much in the domain of securities filing is mere check the box drudgery, with little strategic thought. All a business wants to do is achieve compliance. Such desultorily collected data is nevertheless helpful—perhaps even more helpful. All the disclosure team wants to do is get the disclosure right and then to move on. There is no gamesmanship as to how many sales happened in a certain country because all that matters is accuracy. As already discussed, there is almost certainly gamesmanship happening with the sales factor. It is possible the tax gaming could infect the securities filings, but the opposite strikes me as likely—or more likely—to be true.

And this is not just an intuition. There is a compelling argument that frictions caused by interactions with other domains are already limiting the amount of tax gaming. As of July 2006, there is Financial Accounting Standards Board (“FASB”) Interpretation No. 48 (FIN 48). FIN 48 requires all public companies and some private companies to take into account tax uncertainty by holding reserves. For positions more likely than not to pass muster (over 50%), the taxpayer must assign some probability to the chance of failure (say 40%) and multiply that percentage by the maximum tax benefit. The resulting number must be recorded as a FIN 48 liability. Any tax benefit less likely to succeed than not is to be recorded as entirely a liability. Thus, as one leading practitioner put it: “FIN 48 creates a new model: Unless the planning yields a highly certain result, there can be no financial statement recognition for the uncertain planning.” One might expect that FIN 48 has therefore reduced the incentive for taxpayers to engage in tax planning, and there is some evidence to that effect. There is also significant evidence that managers are more sensitive to accounting measures of their performance relative to tax measures.

Or, more prosaically, having tax results pressure the auditing results creates new bureaucratic stressors on top of the substantive problem that distorting the audit results is improper and can bring about its own consequences.

In sum, data collected for other purposes, such as accounting, is typically subject to different incentives. Further, because such evidence is often produced by other bureaucracies for other purposes, it can be more difficult to manipulate for tax purposes.

D. An Example That Has It All (Mostly)

A recent Maine decision from its Supreme Judicial Court elegantly illustrates these problems and the solutions I am proposing. The taxpayer in this case, Express Scripts, provides prescription management services to intermediaries like insurance companies by providing medications to their customers’ members (such as insured individuals). Consistent with the Maine statutes and the underlying theory of the sales factor, in 2011 the taxpayer sourced sales to Maine on a market basis, that is, sales of its service were apportioned “to the state in which the prescription drug is dispensed to members by the retail pharmacies.”

This sourcing was consistent with the purpose of the sales factor for a few reasons. First, the sales factor is supposed to capture the market provided by the state, and sourcing based on deliveries of the medication is the closest approximation of the market. Second, the sales factor is supposed to be a relatively efficient choice of factor because consumers are relatively immobile relative to property or payroll. Third, permitting sourcing to intermediaries rather than consumers is not only inaccurate, but it is also an invitation to gaming, as was the case here. This is not to say that headquarters are necessarily all that mobile once established, but it is certainly the case that major taxpayers extract tax concessions from jurisdictions all the time in return for locating in the state. Further, a headquarters is just one example of an intermediary. Once the ultimate destination rule is deviated from, then sales can be sourced to warehouses or to wholesalers in a low or no tax state.

So, the taxpayer in this case could and did source it sales consistent with the purpose of the sales factor in 2011. However, in 2012 and following, the taxpayer shifted to sourcing sales to the location of the headquarters of its intermediaries, which resulted in a fourfold reduction in the apportionment factor to Maine. By 2013, the taxpayer had reduced its factor by another third and even asked for a refund of about $1mn for 2011. The taxpayer was assessed additional tax due of about $900,000.

These are large numbers and for just one taxpayer in a small state. It is hard to believe this is the rare case where a sophisticated interstate taxpayer strategically sourced its sales. Indeed, this is the very strategy that a lobbyist for New Jersey argued would proliferate if the state did not make it easier to shift income offshore.

The Maine Supreme Court, in rejecting the taxpayer’s contention that the market for its service was in the states where its clients/intermediaries were located, looked, among other places, to the taxpayer’s securities filings: “In its Forms 10-K for 2011, 2012, and 2013, Express Scripts explained that, pursuant to its contract with one client, ‘Express Scripts provides [PBM] services to members of the affiliated health plans of [its clients].’”

Why was it appropriate for the court to credit, indeed highlight, this evidence? For one, this evidence was not prepared in the context of tax planning or litigation. Further, under the securities laws a non-misleading picture of the material functioning of the business must be provided, subject to substantial penalties. Suppose I am wondering whether to invest in a business providing prescription management services; does an investor want to know how well, and where, it is servicing the headquarters of the insurance companies with which it contracts, or does the investor want to know how well, and where, it services end users? Does an investor want to know if there is competition for similar services in the area of the intermediaries’ headquarters or where the ultimate services are delivered? There are thus substantive reasons—not just bureaucratic ones—to look to this other domain for guidance as to how to improve formulaic results.

III. Applications

In the following sections, I will consider how superior formulas might be actualized identifying probative information collected for other purposes.

A. Business Taxes

In considering the need for better state-level formulas, I have already identified that the main need relates to the sales factor. I have also generally explained what it is that we are looking for in terms of a better formula. Yet these considerations are still fairly abstract unless, as in the Maine case, a taxpayer describes their business model for tax purposes in a manner not consistent with its description for purposes of securities law. Fortunately, there is a fuller model of my proposal provided by worldwide combined reporting (WWCR).

This is not the place to argue for WWCR as a policy matter, though I think that it is an excellent idea. I will note here that one of the primary objections to WWCR has been—and still is—that it would be administratively burdensome. As a matter of public record rather than anecdote, information was collected about this concern during 1980s as part of the litigation regarding WWCR and the result was clear. There was not much of a burden involved in a big corporation providing reasonable estimates as to its foreign operations, which was all that was (and would be) required. Still, things have changed since the 1980s and it is instructive to observe that WWCR would likely be even easier to comply with now.

B. The Ironically Easy Case of Worldwide Combined Reporting

There are two aspects of a multi-national corporation’s (“MNC’s”) foreign operations that are needed for reasonably accurate WWCR: global income and sales for the sale factor. (Note that back in the 1980s MNCs would also have had to calculate payroll and property factors.) Our focus ultimately will be on the sales factor, but we will start with income.

1. Income

There are multiple estimates of foreign income from which a taxpayer can start. Most obviously, there will often be foreign income tax returns. To be sure, these are not standardized, which is one reason why the OECD Pillars and the new Corporate Alternative Minimum Tax start from financial accounting statements. Such information is generally the starting point for WWCR.

Note that there would be a harder issue here if what was needed was accurate country-by-country reporting of income. As we will see in a moment, this is not necessarily how the financial statements are organized, but these statements do of course include total income (and very likely U.S. income), which is all that is needed for a starting point.

There are other sources of useful information on foreign income. For example, for federal tax purposes, there is the information that MNCs subject to Global Intangible Low-Taxed Income (“GILTI”) must report as to the income of their U.S.-controlled foreign corporations “CFCs”).

2. Sales

Just as a business had best keep track of income for its own purposes, so too it had best keep track of sales for its own purposes. That said, there are numerous ways a business can transform the location of sales strategically, and this has apparently been happening with the domestic sales factor. One way to do this, as already noted, is to use intermediaries to source sales in low tax states (or nations). Other strategies include treating lower profit subsidiaries as unitary because that reduces overall income or attempting to isolate profitable entities and rely on federal Public Law (“PL”) 86-272 so that the profits are not taxable. PL 86-272 prevents states from imposing corporate net income taxes on entities whose activities in a state are limited to soliciting orders for sales of tangible personal property approved and fulfilled from outside the state.

Because of the shift to overweighting the sales factor, or using only the sales factor, these manipulations are particularly significant. It should also be acknowledged that sourcing many products and services in the current economy is objectively challenging. As a result, some sales factor outcomes likely also flow from generally compliant taxpayers making good faith and reasonable decisions in the context of their particular circumstances, albeit with potential advantages for the taxpayers themselves. Such taxpayers would comply if given better tools.

As with income, there are relevant sources of information collected about sales for other purposes which might be used to combat tax planning or good faith choices that nevertheless skew the sales factor significantly in aggregate. Specifically, accounting standards already require geographic reporting:

This Statement requires disclosure of information about both revenues and assets by geographic area. Analysts said that information about revenues from customers in different geographic areas assists them in understanding concentrations of risks due to negative changes in economic conditions and prospects for growth due to positive economic changes. They said that information about assets located in different areas assists them in understanding concentrations of risks (for example, political risks such as expropriation).

The SEC expects compliance with this standard and at least sometimes reviews corporate disclosures to assure compliance. Indeed, the information the SEC expects to see is arguably more granular, as it requires disclosures at the national level, at least if revenues are material:

Information about geographic areas is also required to be disclosed based on countries, both the country of domicile and for foreign countries. If a registrant manages its business by geographic regions and determines its reportable segments accordingly, it still must provide the separate geographic disclosures for each country in which revenues are material. Some registrants provide this disclosure by presenting material countries separately within the subtotals by region.

Here is what this disclosure looks like in Amazon’s 2022 Annual Report:

Once one has U.S. sales as a percentage of global sales, all that is left is ordinary domestic sales factor calculation.

The securities law backstop is, of course, only of use as to public companies. And even the accounting standards, and in particular FASB Statement No. 131 relating to segment reporting, need not be complied with by private companies. As an analytic point, this does not matter for my example, which is meant to illustrate how such information can be helpful. As noted at the outset, for a company that must already make geographic disclosures, like Amazon, using securities data as either a starting point or check seems obviously warranted and administrable. There might not be similarly easy tools for improving the domestic sales factor, but my point has been to provide a paradigm.

As a practical matter relating to worldwide combined reporting, this fair observation as to the scope of which firms comply with financial accounting should not be overblown. When it comes to income shifting, we are primarily concerned with a relatively small number of very large and profitable multinational corporations, as they are the ones with the supra normal returns and the ones likely to do the shifting. As to these taxpayers, there are three possibilities. Either they are (1) already public, (2) they are private and have collected relevant information as part of their dealings with other business partners (such as investors), or (3) they are private and have not generated this information for some other purpose. If they are in the last category, then, given their size and sophistication, asking for the provision of this information is not likely to be excessively burdensome, but tax authorities will lack the backup function of public reporting. Thus it would make sense for tax authorities to work with these taxpayers to develop bespoke reliable indicators.

The policy upshot as to worldwide combined reporting is thus that, like the corporate alternative minimum tax (“CAMT”) and the Pillars, it would make sense for states to make reporting mandatory only for the largest corporations, for the reasons already discussed. To begin with, as these are firms that represent a significant portion of the tax base, they have the greatest capacity to engage in sophisticated tax planning to strategically shift sales, and they manifestly have a greater capacity to comply than smaller firms. And many such firms are also collecting useful backstop data anyway.

C. Back to the Domestic Sales Factor

Unfortunately, I have not observed state-by-state sales data disclosure as part of financial accounting or federal securities law. One interesting reform to consider is whether some state-level disclosure, or at least disclosure by region, is appropriate, since such disclosures could well be “material.” This could be because certain states, say California, provide markets as large as that of many countries and so should be considered as material as disclosure by country. Certainly disclosure by regions of the country, say the Western United States, could be material as having concentrated risks (fire, drought etc.).

If national financial regulators could not be convinced to mandate such disclosure, then states might be able to require them. It might seem odd for states to mandate disclosure of percentage of sales for securities and tax purposes, but given their different purposes and enforcement mechanisms, it might well be worth it.

There may well be other sources of reliable sales data assembled for other purposes. One option relates to gross receipts for purpose of the sales tax. Sales tax receipts are far from a panacea because so many transactions are not subject to the sales tax (services). Still, for many transactions, sales tax gross proceeds are probably useful.

There are also likely other sources of useful regulatory information. For instance, the California Consumer Privacy Act (CCPA) grants consumers the “Right to Know What Personal Information is Being Collected.” This rule only applies to California residents and hence businesses that collect consumer data, which is presumably every large business at this point, must have some means of identifying California customers. Though clearly somewhat limited in scope, requiring a taxpayer to include all gross receipts associated with California customers for purposes of the CCPA would seem to be a step forward, especially since the California Attorney General has already taken enforcement actions in connection with the CCPA.

More generally, the states can follow the OECD by establishing the general priority of information collected for other purposes. According to the current draft commentary on apportionment for purposes of Pillar One, taxpayers must use an “Enumerated Reliable Indicator,” which is defined, in part, as “the Indicator is relied upon by the Covered Group for commercial purposes or to fulfill legal, regulatory, or other related obligations.”

This is a step forward relative to current U.S. practice. Consider the Model regulations of the Multistate Tax Commission (“MTC”), which generally call only for using the “books and records kept in the normal course of business.” The OECD model appears to indicate that reliable methods are those that rely on some kind of third-party verification, not just on whether the records are ordinarily maintained. For instance, either the method is relied on for some commercial purpose—such as number of sales for a commission—or it is used in some other regulatory regime. If such a reliable method is not available, then another method is possible but is subject to “advance certainty review” to ensure that the method “produces results that are consistent with the revenue sourcing rule for the category of Revenues at issue.” It would seem wise to make it clear to taxpayers in the U.S. that their homegrown estimates are not, ordinarily, reliable without more.

1. Use More Refined Data.

When all else fails, both the states and the OECD allow for default rules based on statistical data. In the U.S., this typically means share of U.S. population. Before moving on to how to improve the default, it is worth spelling out why this is important.

It has already been shown that if states deviate from the ultimate destination rule for sales then they expose themselves to the kind of gaming illustrated in the case from Maine. What if a state has the correct rule? The rule is not self-actualizing, and it is surely true that in many cases it is harder for taxpayers to locate their ultimate consumers. In some cases, especially when it is likely helpful, taxpayers must be sorely tempted to just use the default.

Here is an example illustrating the problem from the perspective of states. Suppose two firms each serve half of the California market for some good or service. As to one firm, from an omniscient view, California represents 7% of its sales; for the other, the percentage is 21%. If averaged together, then the apportioned share of these two firms’ income to California would be 14%, about California’s share of GDP. In both cases, however, it is not so easy to figure out the location of the ultimate consumer. The first firm, with an apportionment percentage of 7%, less than California’s share of the U.S. population (11%), has incentive to spend the time to substantiate a sales factor of 7%. The second firm, which knows it makes a disproportionate amount of its sales to California, does not. It is happy just to use 11%. Accordingly, in this scenario, California ends up with an overall 9% apportionment factor. And the taxpayer who drove this result did not engage in particularly aggressive tax planning.

Hence, it would be better to have a more accurate default than population. I think share of GDP would be better because that seems better correlated with the size of the market provided. The OECD, in its model allocation regulations for Pillar One, goes farther. The OECD permits the use of (somewhat) industry specific “allocation keys,” much like the states have special industry apportionment formulas. For example, there is a “Passenger Air Transport Allocation Key” and a “Service Allocation Key.” But there are some more general keys available for use in certain situations, like the “regional allocation key,” which

means that Revenues are treated as arising in Jurisdictions, provided they are in the Region, in proportion to the percentage of their share of the final consumption expenditure for the most recent calendar year that does not end after the Period ends expressed at current USD prices as published by the United Nations, and if not available, the value in current USD as published by the World Bank and converted to EUR at the Average Exchange Rate.

There are two benefits to this approach that should be noted relative to defaulting to population. First, it relies on independent data and, second, this key is generally more refined than population and so more likely to be a reasonable approximation. Indeed, to the extent the OECD keys require use of population, major taxpayers have, rightfully, protested to the OECD that headcount is less accurate than the use of final consumption expenditures, or failing that, percent of GDP.

But the OECD uses UN information that is more precise than GDP; can the states do so as well? The answer is yes. It is worth a quick consideration of what is available.

The Economic Census provides data on the value of sales by industry and state every five years. Alas, as MTC economist Elliot Dubin explains, these sales are by origin of the sale and not destination. However, this data can be broken down by industry subsector, such as “Hotels.” Hotels provide a service that, by definition, is not likely to cross state lines, and so the Economic Census could provide a reasonable approximation of a sales factor for a multistate corporation such as Hilton. For instance, relying on the Economic Census, California had 14.73% of all hotel sales in 2012, as compared to its 13.23% share of GDP.

This data will only go so far because most of the sales we are interested in will be interstate. As a result, Dubin used a more involved procedure for all industries. This procedure was first developed by the U.S. Advisory Commission on Intergovernmental Relations, then used again by the Tax Policy Center.

The Bureau of Economic Analysis (BEA) tracks the input and output of industries. For some industries the output is the end user. For other industries, the output is to some other industry. In order to come up with an apportionment percentage, one needs to assess how many end users are in a given state if the product is consumed. Alternatively, one needs to assess how much industry is in a state if an input is to be used in business. As for final users, this methodology assumes that the final users track the share of that state’s GDP. Note that this measure can be refined insofar as we have some data on particular consumption patterns per state. As for intermediate users, we have data from the BEA on the state-by-state distribution of various industries.

Two examples will hopefully make this procedure clearer. First, take food and beverage purchases. According to the BEA, over 99% of purchases from food and beverage stores are for final consumption. This is a category for which we do not need to worry about intermediaries. We could sensibly say that consumption for this industry should track state GDP and so California’s factor should be 13.23% and Alabama’s 1.17% for 2012. However, using the BEA’s personal expenditure data for 2012, we get a 1.40% factor for Alabama and an 11.92% factor for California, suggesting that the BEA data may permit a more reliable view of final user consumption.

What about an intermediate good, like fabricated metal? According to the BEA, 16.52% of fabricated metal is ultimately used in the construction industry. Going back to the 2012 Economic Census, the value of construction in Alabama was 1.31% of the national total, while the value in California was 10.87% of the national total. Therefore, in putting together a model apportionment factor for fabricated metal, about 16% of sales should be overweighted (related to GDP) as to Alabama and underweighted as to California.

Three further observations should be made. First, to the extent that the BEA information could be more useful, states should not treat the data provided as fixed, but should lobby for the creation of data useful for state tax administration purposes. Second, to the extent that making this information into a meaningful key requires work—and should be standardized—this would be an appropriate task for the MTC or even a state revenue agency. At least one earlier draft suggested that the OECD would take on a similar role in some circumstances.

Finally, reasonable defaults can be used as an information gathering device. Again, in current practice, a taxpayer investigates its operations and then provides an apportionment factor (or uses a default, typically population). Given the evidence that firms are self reporting factors far lower than the current default (population), much less a better default, there is reason to believe there is gaming occurring. But suppose there was a presumption that the default (say GDP) was correct and that the taxpayer needed to provide evidence otherwise. The standard to overcome the presumption need not be high and many taxpayers could (properly) overcome it. Still, the requirement to come forth with one’s analysis and evidence in an organized way likely creates a new friction for taxpayers more weakly inclined to gamesmanship.

2. Knock-out Rules.

But why should a taxpayer use global GDP in its denominator when it primarily makes sales to Europe? A similar question can be asked for the U.S. Why use U.S. GDP for a taxpayer that makes most of its sales on the west coast?

Accordingly, the OECD model requires that when using an allocation key, a jurisdiction must be excluded “where there is a legal, regulatory or commercial reason such that it can reasonably be concluded that Revenues did not arise in that Jurisdiction or group of Jurisdictions.” In other words, if a firm sells all its goods in Europe, it cannot dilute its profits allocated to Europe by including the U.S. economy in the relevant allocation key.

In the integrated U.S. economy, this rule might need some further refinement. For example, for a big multistate corporation, it is quite likely there will be some sales in almost all states. Perhaps the knock-out rule would then apply as to states in which de minimis sales were made (say less then 1% of total), but then some factor dilution should be permitted, say 10% to account for all knocked-out sales, subject to alternative apportionment.

3. Ordering.

A further refinement suggested by the OECD model has to do with when the allocation keys can be used. In a few instances, the keys must be used by the taxpayer, but the usual rule is that the taxpayer can only use a key if it “demonstrates that it has taken Reasonable Steps to identify an Enumerated Reliable Indicator.” In other words, as least in theory, a taxpayer cannot just use the default rule if it suspects that it will provide a better result than using a reliable indicator.

D. Special Issues Relating to Digital Products and Taxes

The rise of the digital economy has made locating sales harder but has also raised the question of whether sales should be the only proxy for activity in the jurisdiction. For example, India was considering a proposal to introduce formulary apportionment, including a fourth factor measuring “user intensity” for businesses in which user participation is key.

Many jurisdictions around the world have adopted entirely new digital levies, which also raise new questions as to apportionment. Maryland has the U.S.’s only standalone digital advertisement tax (for now, anyway). The Maryland tax is apportioned, as it must be, so that it taxes only the gross receipts arising from transactions in Maryland. The regulations implementing apportionment explain that digital advertising revenues are derived in Maryland “when any portion of those services are accessed through a device located within the State.” The choice to focus on devices, rather than users, is controversial. As I understand it, the UK focuses on the users for its digital services tax, while France focuses on the device.

As of now, I am agnostic as to the best approach to apportion digital taxes or how to calculate a new digital apportionment factor, but I am not agnostic on the need for more thought as to the best approach. As to the best approach, I should note that here, too, there is the possibility of piggybacking on other regimes, such as the CCPA, and that would be ideal as we are looking for stable, long-term, incentive-compatible and administrable solutions.

It is also worth noting that there is a certain strategic unreality to some of the critiques of the apportionment methods for digital services, which should remind us of the supposed allure of transfer pricing as opposed to formulas. For instance, it is objected that IP addresses do not reliably provide information on a user’s location in every case. I cannot assess this but suppose this is true; how important is this fact? Apportionment got started—and still generally uses—track or road miles in a jurisdiction for transportation industries. Is the contention that these formulas for digital location are significantly less reasonable than that? Or, is the worry that users of these digital services will for some reason systematically collude to increase (or decrease?) a taxpayer’s sales that would be apportioned to a jurisdiction. More concretely, why would imperfections in Maryland’s approach not get washed out, with some Maryland devices (or users if another formula were used) not showing up as located in Maryland and vice versa? What is it about Maryland such that users would collude to appear to be in Maryland in a manner that would unreasonably increase sales attributed to the state? And what better approximation is available? If there is one, then I am certainly for states considering it.

E. Conclusion on Business Tax Apportionment

The main point of this section is that should states wish to tax mobile capital in the form of business profits more effectively (and they should), then reforming their apportionment formulas is one important and promising avenue. States that use modified gross receipts taxes would also benefit from these improved formulas for locating sales. Until recently, the improvement of apportionment formulas was a wholly domestic dialogue. However, this is now an international discussion, and the learning can go both ways. In particular, utilizing information collected for some other purpose and, ideally, subject to another regulatory regime should be utilized as a default more than is the case under current U.S. practice.

A final point. Because I have argued that the primary need for reform relates to sales systematically not being sourced properly, especially to higher tax states, my focus has been on how these reforms would help states. They would also help taxpayers in several ways. First, in particular cases, the use of a more refined formula will (appropriately) reduce the sales factor in some situations. Second, and in all cases, more granular guidance allows taxpayers that want to comply the ability to do so more quickly, efficiently and securely.

IV. Apportionment for Individuals

The notion of residence has long been binary—one is a resident of only one country, state or city etc., and not of any other jurisdiction. So too the notion of source for individual earned income; one works in one place. Sometimes, as when people move within a year, residence is apportioned. Source is also sometimes divided, as for the income of professional athletes. But these are exceptional cases. The regular rule for individuals is binary, which has significant administrative benefits and, happily, has conformed, roughly, with how many, or most, people live.

Yet even this happy story was undermined somewhat by the operation of the credit system to prevent double taxation. If a source state, say New York, had a tax rate that was as high or higher as that of a resident state, say New Jersey, then the resident state would get little or no income. Maryland tried to at least ensure a certain amount of income tax would remain in-state for residents with income out-of-state, but the Supreme Court struck down that approach in Wynne.

Recent developments, particularly the rapid increase of work from home (WFH), has put even more pressure on the residence, source, and credit model. Which state is the source state if I work from home in New Jersey three days/week and commute to New York two days? What if I almost never commute into the office? What if I never commute? WFH might not be as big a phenomenon as some suggest,, especially for most workers, but it has clearly changed work patterns for many and, as just explained, the current system was hardly perfect.

Some prominent proposals seek to recreate the binary of source and residence by fiat, through establishing a bright-line rule, typically based on days of physical presence. Such proposals recognize the issue as to a growing class of workers but then in effect doubles down on a rough solution (binaries) based on a factor that cannot bear this much analytic weight (physical presence). Rather, I think we need new rules to address a new situation more appropriately.

What would such rules look like? In many cases, there should be reliable records kept in the ordinary course of business (e.g., all workers come into the main office two days/week). Of course, depending on the context, such information might not be available or result in a reasonable split. Suppose a law firm has one office in New York City and a partner in New Jersey who never commutes into the office. Would NY have zero claim as a source of that lawyer’s income? To account for such situations, there could be rules of thumb based on the location of the primary nodes of the business. Perhaps in the one office scenario, half of the partner’s income would be sourced to New York.

But what if the business is larger and more complicated, with offices all over the country? Now we would need some means of apportioning individual income based on several variables, such as nodes of the business and how often the employee comes to a physical space. One might also use a kind of knock-out rule as to remote employees if most of the other employees can be reasonably located. That is, if a firm has 100 employees, with 90 evenly distributed in two states, then this is a firm whose employees have source income in two states and the 10 fully remote workers should have their revenue sourced to one of the two permanent offices, at least in part. Again, one might have some default discount for fully remote workers, say no more than 50% of their income can be sourced to the combination of the source states.

But what if the offices are of different sizes? Perhaps then one could use the relative number of employees in each office. What if the offices do very different things? Or what if a jurisdiction is interested in capturing its locational advantage in the formula? Consider New York and finance. In those cases, a state might look to the North American Industry Classification System (NAICS) code of the business and/or how the work of the employee ought to be categorized under such a code. The use of NAICS code for taxing businesses exists, as demonstrated by Nevada’s Commerce Tax and San Francisco’s gross receipts tax, both of which apply different tax rates depending on the NAICS code of the business.

Yet the use of such new formulas does not take advantage of the kind of complementary data we were looking for in the context of business apportionment. This is not to say that some information generated for other uses might not be available. There could be card swipes maintained for security purposes or perhaps even insurance contracts that could be probative. Firms might also be presumed to rent/own space that is the right size for the workforce they expect in a given location. Consider a New York law firm that understands much of its workforce will not be in the office at least part of the time. If the office has 300 nominal employees, then perhaps it will rent a space big enough for 150 because it is expecting half of the employees to be present at any given time; a large number of offices will therefore not be assigned but be available through reservations. If the various employees in the aggregate only report the equivalent of 75 employees using the office, that is half of what the office is designed to house, then this is prima facie evidence of a systematic misfire. Why would the law firm pay for twice the space that it needs?

There are also statistical estimates of employment by state by industry, perhaps subject to adjustment by locational advantage. Perhaps knock-out rules could make these formulaic tests more robust too—say, if a jurisdiction does not contain at least 5% of a firm’s workforce but other jurisdictions do contain sufficient clusters (20%), then the jurisdictions that do not meet the threshold are knocked out (at least as a source for the individual’s income).

There is also an emerging securities disclosure regime regarding human capital. Such disclosures have been required since 2020, with about ¼ of disclosures already providing information as to geography. The SEC recently solicited comments on expanding these disclosures. It is not clear whether geographical disclosures will ultimately be required or would even be useful, but they might be. The underlying premise is that for many firms their human capital is their main asset, so knowing where these workers are located can be material.

Formulating a new methodology to address the issue of sourcing the income of (more) remote workers does seem like a lot of trouble. The simple binary seems easier. However, once a reasonable formula and methodology is chosen, it ought not be too onerous. After all, the complexity of the calculation will tend to scale with the size of the business so that smaller businesses should have relatively simple calculations.

But why exert any effort? Partially, this is a matter of politics. There needs to be some way of splitting the pie that does not cause too much fiscal harm to states (and localities) that contribute to the income production of individuals. As a matter of retail politics, one ought to expect a source state, such as New York, to look after its fiscal base. Thus, if one acknowledges that “convenience of the employer” type rules are not a great fit in the modern economy (as I do)—but one thinks they are constitutional (as I do)—then one must come up with some other broadly fair and roughly revenue neutral alternative. One might wonder how a shift to apportionment could possibly even out for a state like New York. This is an empirical question but it is worth noting that New York is also a desirable place to live (for some) and there are a certain number of taxpayers who live in New York but have income sourced outside of New York.

But there are larger stakes if one believes, again as I do, that aggregations still matter. If New York is providing valuable aggregations, say in finance, then this is of benefit to many, not just those who earn a premium for working for a New York firm. It is therefore in the national interest generally to arrive at a reasonable solution.

V. Intangible Apportionment

Judging by the space intangible apportionment takes up in tax publications such as Tax Notes State, the apportionment of intangibles is a major issue. Certainly it seems to be a big planning issue, though, as argued below, sound policy seems clear (use apportionment), as does the law (permitting apportionment). The underlying issue is caused by the same source/residence binary as just discussed for individuals. As already explained, the income of a multistate business is apportioned among the states in which it does business. In other words, this income is taxed on the source principle. Also, as a general rule, the income a taxpayer earns from the sale of intangibles—say a stock—is sourced to where the taxpayer is a resident.

And thus, in the typical case driving discussion of the issue, a taxpayer builds up a profitable business in one state, a state with an income tax, and then sells their shares of that business in another state, a state without an income tax. As a policy matter, it seems clear that the first state has a claim to an apportioned share of the capital gain generated by the sale of the business. As shown in the recent debate about this in connection with the VAS Holdings case, in the end, thoughtful critics of this decision accept that there is not a meaningful economic difference between ordinary income and capital gains income.

The strongest policy counter to apportioning capital gains is that if some states still use the residency binary then this could lead to double taxation if the taxpayer has moved from California to, say, New York. Yet even if this were a common problem, then there are policy answers from adjusting when resident states give credits to apportioning capital gains. I will discuss this latter possibility below. There is no reason for the source state to accept getting nothing. To the contrary, given the number of states that would not tax the gain at all, such rules are just an invitation to create nowhere capital gains income.

The law permits apportionment in this case. As a general matter, relying on the usual apportionment formula passes muster. There is also no problem with a state taxing an apportioned share of source income going to investors out of state, nor is the common all-or-nothing residence rule constitutionally required for intangibles. The counter argument relies on U.S. Supreme Court dicta and is not persuasive.

Note that for this typical case of a founder selling shares of their business in a no-tax state the legal issue is whether or not to apportion at all versus how to apportion. Thus these cases do not really raise the need for new formulas. Hopefully, the improved formulas discussed above could be used. For example, if we are sourcing intangibles reasonably well for ordinary business tax purposes then this methodology could be used (and in effect would be) for sourcing the capital gains derived from intangibles. California’s Franchise Tax Board has recently followed a similar logic as to the sale of partnership interests.

There is, however, a need to think about formulas when the income from a sale of intangibles is more tenuously related to business income. In such case, we cannot rely on the underlying apportionment factors of the business.

Consider a taxpayer selling a portfolio of stocks that gained value during the years a taxpayer lived in the state. As a matter of logic and law, it would seem like the state of residence for the accumulation should have a claim, and it does under current law. Yet the current resident/no resident binary rule leads to peculiar results in some cases. If I move to California with a lot of built-in gain and sell my assets, then California, as the resident state, taxes 100% of my gain. On the other hand, if I have lived in California my whole life—and my fortune has accumulated there—then I need not pay any tax on the gain if I sell my assets after I leave California. It is hard to come up with a principled approach to residency-based taxation that says that California has a claim on 100% in the first case but 0% in the second.

No doubt the main argument for this rule is administrative, and more complicated rules applicable to all taxpayers would be difficult, but applying more refined rules to a small set of very wealthy taxpayers seems quite possible. Indeed, there are models for dealing with situations like this in current law. Consider deferred compensation. When an employee exercises the option they received as compensation, even years later and living in a different state, they owe tax on the value of the compensation in the state in which it was earned. But how much of the value of the stock was earned in a state if the employee earned the stock while working in more than one state? One general approach is a timing rule. The value of the compensation is apportioned based on how much time was spent in each state.

Something similar can be done with the appreciation of intangibles—at least in certain cases. For a given asset, gain can be apportioned based on length of residence. To be sure, such a regime is imprecise and there are administrability challenges as taxpayers move. Other options include solving the problem through taxing economic income—or some portion of it—when earned and hence some kind of mark-to-market system. Note that solving the problem of mark-to-market reform likely also involves the use of formulas for valuing assets and/or deferred tax.

A more limited version of such a reform would have borrowing against appreciated gain trigger tax. There could also be a choice of options: pay 50% of economic income now or pay 100% of apportioned income later. Again, it would make sense for any such rules only to kick in at high value/income thresholds and be subject to the possibility of alternative apportionment. If a taxpayer can make a strong showing as to where the appreciation occurred then such a showing should plausibly carry the day, but it seems burdensome and to invite gaming for this to be the usual rule. There would presumably need to be anti-abuse rules.

The alternative to engaging in this complexity is to permit old rules to continue to distort choices and cost states revenue from those most able to pay—but in many cases least likely to.

VI. Conclusion

Though this Article does not include a menu of superior formulas for state tax purposes, it has outlined how such formulas may be developed. The use of formulas should also expand to other areas, especially if the formulas are improved. This Article has attempted to counter suggestions that there are rigorous sourcing techniques superior to formulas. Most importantly, this Article has aimed to counter the argument that there is nothing to be done as to the limits of the current situation. It is true that, as things stand, flawed formulas (or the lack of even an attempt to use formulas) often serves to benefit systematically the most profitable corporations and wealthiest taxpayers in rough proportion to their willingness to engage in tax planning. But this is not how things have to be; this is how we have allowed them to be. There are better data available from a number of sources that could improve apportionment formulas and result in more equitable tax reporting for taxpayers and taxing jurisdictions alike.