Although the treatment of financial institutions and services is often said to pose a difficult issue for consumption taxes, that statement is misleading. As discussed below, this issue poses challenges for both income and consumption taxes. The problems are more visible, but no more severe, under consumption taxation.
The financial sector is a significant part of the economy. The national income and product accounts data reported by Harris et al. (2011) show that the finance and insurance industry had value added of $1,172 billion, or 8.3 percent of gross domestic product (GDP), in 2009. Of this amount, $568 billion was employee compensation, $45 billion was taxes net of subsidies, and $558 billion was gross operating surplus (business cash flow). The industry had 5.8 million employees, 4.2 percent of employment for all industries. The sector includes a diverse set of institutions. Of the total 2009 value added, $514 billion was in "Federal Reserve Banks, credit intermediation, and related activities," $175 billion was in "securities, commodity contracts, and investments," $424 billion was in "insurance carriers and related activities," and $58 billion was in "funds, trusts, and other financial vehicles."
There is little dispute about the treatment of financial services that are provided to firms for use in the production of real output. Any tax imposed on such services should be offset by a deduction for the purchasing firm. So long as such purchases contribute (on the margin) to the production of real business output, the tax imposed on that output is sufficient; the imposition of a second tax would give rise to inefficient production. A tax deduction for business purchases of financial services has the same economic justification as a tax deduction for business purchases of office supplies, machinery, or raw materials.
This justification vanishes for consumer purchases of financial services. Just as consumer purchases of consumer goods are not exempted, neither should consumer purchases of financial services.14
It is easy to see this in a simple example. Consider an economy in which Smith produces ten apples today, and Jones produces ten apples tomorrow. Because both workers wish to eat some apples today and some tomorrow, it is beneficial for Jones to borrow some apples produced by Smith today and to repay Smith from the apples that he will produce tomorrow. Suppose, though, that the two cannot make this transaction happen unless a third person, Thompson, puts in an hour of labor to make the arrangements. A sales tax, VAT, or X tax must separately include the financial services in the tax base. The tax on the apple production taxes Smith's and Jones's work, but not Thompson's work.
The X tax must therefore distinguish between personal and business use of financial services, just as it must distinguish between personal and business use of cars, computers, and meals. But it is worth mentioning what personal and business use means. The relevant question is whether the financial service pays for itself on the margin in terms of additional real business output on which the X tax is imposed. This may not always be the same question as whether the services are provided to households or to firms. If financial services purchased by a firm facilitate its transactions with its owners without affecting the production of real output, then the services are personal in nature. If financial services purchased by a household lead to improved resource allocation that increases real output, then the services are business in nature. For administrative reasons, though, the dividing line will probably have to be drawn on the basis of whether the purchaser is a household or a firm.
The tax challenge posed by financial intermediation is the need to look behind the labeling of transactions to determine whether they are actually real or financial in nature.
Different Ways to Charge for Financial Services. Financial transactions would pose no special problem, for either income or consumption taxes, if financial institutions charged explicit fees for all of their services. In that case, the costs of financial services would be readily identifiable, allowing them to be taxed in the same manner as other services, with the only challenge being the need to distinguish between business and personal services. In practice, however, financial institutions often recover the costs of providing services by paying below-market, or charging above-market, interest rates rather than by charging explicit fees. This arrangement makes it difficult to identify the costs of financial services and properly tax them, because the costs look like variations in interest income or expense, which warrant different tax treatment.
Suppose that a 5 percent interest rate is offered on safe assets that provide no financial services, such as Treasury bills. A checking account depositor with a $1,000 balance will not receive a $50 net annual return, even though she would have done so if she had held Treasury bills. The reason is that the depositor receives, and must pay for, costly services provided by the bank, such as check processing and record keeping. If the services cost $30 per year, the bank must charge the depositor these costs in some form, allowing her to clear only $20 per year on the account. The question is the form in which the charge is made.
The bank could pay a 5 percent interest rate on the account but then charge a $30 explicit fee. Alternatively, the bank could pay a 2 percent interest rate and charge no explicit fee. Or the bank could use a mixture of the two approaches, such as a 3 percent interest rate and a $10 explicit fee. In a no-tax world, the choice between these methods might turn on a variety of considerations. Economic efficiency might be promoted by charging fees closely tied to marginal costs; for example, it may be efficient for check-processing costs to be recovered through a fee based on the number of checks written, rather than through an interest income reduction linked to the account balance. On the other hand, administrative savings might be achieved by simply reducing the interest rate rather than administering a whole set of fees.
A similar process occurs on the other side of the ledger. Suppose the bank lends the $1,000 to a borrower. For the time being, assume that there is no default risk on the loan. If the bank performs $10 of services for the borrower, there is again a choice of ways to charge for these services, including the option of charging a 5 percent interest rate and a $10 fee and the option of charging a 6 percent interest rate and no fee.
In a world with income or consumption taxes, proper tax policy requires that true interest income and expense be treated differently from the costs of check processing and record keeping. In this example, the true interest rate on both the deposit and the loan is 5 percent, regardless of what "interest rate" the bank may quote, because 5 percent is the compensation for the use of money. Similarly, the true costs of the financial services are $30 for the depositor and $10 for the borrower, regardless of the fees that the bank charges.
Assume that the bank's $40 costs of providing services to the depositor and borrower consist of $20 wage payments to bank workers and the purchase of $20 of supplies from another firm. Also assume that the supply firm pays its $20 of receipts to its workers in wages. For simplicity, the bank and the supply firm are assumed to use no capital and to have no business cash flow.
The treatment of this transaction under a conceptually correct income tax, VAT, and X tax is shown (in italics) in the first, third, and fifth columns of table 6-1, under the assumption that the depositor and the borrower are both households rather than firms. In each case, the $40 of financial services is taxed, and there is no net tax on the borrower's payment of true interest to the depositor.
Under a conceptually correct income tax, savers are taxed on interest income, and borrowers deduct interest expense, so taxable income is $50 for the depositor and negative $50 for the borrower. The workers are taxed on their wages. The firms have no tax liability because their sales receipts are offset by wage payments to their workers.
Under a VAT, the true interest income and true interest expense are disregarded because the VAT is a real-based tax system. The bank and the supply firm are taxed on their value added. The X tax is similar to the VAT, except that the wage component of each firm's value added, which is the only component in this example, is taxed to the workers.
All three tax systems readily achieve the correct treatment if the bank applies the 5 percent market interest rate to both the checking account and the loan and charges explicit fees for its services.
Taxing Mislabeled Transactions. But what happens if the bank charges no explicit fees, instead paying a 2 percent interest rate to the depositor and charging the borrower 6 percent? If the tax system naïvely accepts the bank's labeling of the transaction, incorrect taxation occurs, as shown in the second, fourth, and sixth columns of the table.
Under the naïve income tax, the bank and supply workers continue to be taxed on their wages, and the supply firm continues to have no net tax liability. The same is true for the bank; it reports $60 interest income and deducts $20 interest expense, $20 wage costs, and $20 supply purchases. The depositor reports $20 interest income, while the borrower reports negative $60 interest income. The combined tax base is now zero; the financial services have vanished from the tax base. The problem is that the financial services have been disguised as a reduction in interest income and an increase in interest expense, each of which lowers taxable income.
The naïve VAT encounters fundamentally similar problems. As under the correct VAT, there is no tax on workers or on the depositor and the borrower. The supply firm's treatment is also the same as under the correct VAT, as it reports $20 value added from its sale of supplies to the bank. The bank, though, has a value added of negative $20, reflecting the fact that it spent $20 on supplies while appearing to earn nothing from its real business activity. (We assume, for the moment, that the bank can obtain refunds for its negative tax base.) The naïve VAT ignores the bank's $40 net margin on deposits and lending because this spread looks like net interest income. Once again, the net tax base is zero and the $40 of financial services escape tax, thanks to their interest disguise.
The naïve X tax is identical to the naïve VAT, except for the treatment of wages. The tax on the supply firm's activity is imposed on its workers rather than on the firm. The bank's tax base moves even further into negative territory, to negative $40, as it deducts its $20 wage payment on which its workers are taxed. The net tax base is still zero, and the financial services still escape tax.
Comparison of the Taxes. There is no net tax on the financial services under any of the naïve taxes because none of them recognizes that some of the "interest" flows are actually real transactions in disguise. Under the income tax, the flows are excluded from the tax base through offsetting interest income tax and interest expense deduction; under the VAT and X tax, the flows are simply ignored. To be sure, these tax systems yield the correct outcome if both the depositor and the borrower are businesses, as there should then be no net tax.
The fact that the same fundamental outcome occurs under each of the three tax systems demonstrates that the problem of improper taxation of financial services is not unique to the X tax or to consumption taxes. As Bradford (1996a) notes, the only difference is that the problem is more visible under the VAT, where it is reflected in the bank's negative tax base, than it is under the income tax, where it is reflected in incorrect measures of the depositor's interest income and the borrower's interest expense. As shown in the table, the problem is even more visible under the X tax than under the VAT, because the wage deduction makes the bank's tax base even more negative.15
This problem of improper taxation of financial services is not confined to banks, as many other financial institutions also provide services for which explicit fees are not charged. For example, insurance companies often cover the costs of such services by charging premiums in excess of benefit payments, and brokers often recover the costs of their services from the bid-ask spread. As we discuss below, any solution to the problem must apply to a wide range of financial institutions.
A number of solutions have been proposed, some of which are more promising than others.
Criteria. Any solution to the problem should have the following properties:
As highlighted by the italicized terms in the above bullet points, we omit three potential criteria from our list. First, we do not insist on a unified tax system for financial institutions and other firms; that goal cannot be achieved because the economics of the problem require distinctive treatment of financial institutions.
Second, we allow financial institutions to be subject to an unfamiliar or counterintuitive tax system, an outcome that occurs under the method we ultimately propose. As explained below, we view that as an acceptable price for satisfying the other criteria. But we shield ordinary nonfinancial firms from such problems.
Third, we do not require that the tax imposed on risky transactions always equal the value of the financial services, ex post. By merely demanding equality in present value, which is sufficient to provide the correct incentives, we gain enormous flexibility, as discussed below.
We now discuss two proposed solutions that we have opted not to embrace. We can quickly dismiss one approach, which simply exempts banks and other financial institutions from tax. At first glance, that method may seem appealing. In table 6-1, setting the bank's tax liability to zero moves halfway to the correct outcome under the VAT and actually yields the correct outcome under the X tax. (In both cases, exemption is a tax increase, as the bank's tax liability would otherwise be negative.) Moreover, the method spares the bank from complicated tax computations, indeed from any tax computations at all. But this approach fails to distinguish between the provision of financial services to households and the provision of such services to firms, as it would bring both sets of services into the tax base. That outcome would be improper when the services are provided to firms, given that there would be no easy way to provide the firms with an offsetting deduction under this method.16
The other method that we have rejected requires much more elaborate discussion.
Basic Interest-Spread Method. The interest-spread method can be used under an income tax, X tax (or flat tax), or VAT (or retail sales tax). Hall and Rabushka (1995, 73–75) and the FairTax plan embrace this method. As explained below, the national income accounts also use the spread method to measure the aggregate provision of financial services.
As we will see, the method has some appeal, along with some limitations. When it can be implemented successfully, the method directly solves the problem by correcting the mislabeling of the costs of financial services.
The method imputes an interest rate on each financial transaction and uses it to distinguish true interest income and expense from the cost of financial services. In the above example, the conceptually correct 5 percent interest rate is imputed to both the checking account and the loan, deeming the borrower to pay, and the depositor to receive, $50 of true interest. The spread method recognizes the difference between the stated interest and the imputed true interest as the cost of providing financial services and taxes it as such. The bank is deemed to charge the depositor $30 and the borrower $10 for financial services and is required to treat these amounts as receipts from the sale of real services, even though it has labeled them as interest.
Adjusting the Spread Method for Risk. Unfortunately, this method has great difficulty specifying the proper interest rate whenever risk is present. In the above example, the checking account can be treated as riskless, because the bank has a definite obligation to furnish the depositor the account balance on demand, and it is virtually certain that the bank (or a deposit insurer) will fulfill this obligation. In our description above, we also treated the loan as riskless by assuming that it was fixed-rate and that default could not occur. In practice, however, the loan would surely be risky. If it is an adjustable-rate loan or if there is a prepayment option, the amount and timing of the borrower's repayment are uncertain. Of course, there is also the risk that the borrower will default. It would be particularly misguided to overlook default risk in the wake of the 2008 financial crisis. Surprisingly, however, much of the literature on the spread method does ignore default risk.
A simple example suffices to illustrate the challenges facing the spread method; the challenges become more severe in more complicated examples. Suppose that the loan lasts one year, that the probability that the borrower will completely default by making no payments is one in twenty-one (and the probability that the borrower pays in full is twenty in twenty-one), and that the bank and borrower are risk neutral. Then, the interest rate on the loan, in the absence of any financial services, would be 10.25 percent, rather than the safe rate of 5 percent, to account for the default risk. (The twenty in twenty-one probability of receiving an extra 5.25 percent above the safe rate compensates for the one in twenty-one probability of losing everything, which is 105 percent worse than receiving the safe rate.) If we continue to assume that the borrower receives $10 of financial services and that the bank does not charge any explicit fees, the interest rate on the $1,000 loan is 11.3 percent. Note that the interest rate must be increased by 1.05, rather than 1, percentage points to cover the cost of the financial services, because there is one chance in twenty-one that the interest will not be received.
What happens if the spread method ignores default risk and treats the 5 percent safe interest rate as the true interest rate applicable to this loan? Then, the bank is taxed as if it had supplied $63 of financial services to the borrower, when it actually provided only $10. The entire 6.3- percentage-point excess of the loan's interest rate above the safe rate is treated as covering the cost of providing services, when 5.25 percentage points actually compensate for default risk. To work, the spread method must be modified to account for risk.
In principle, an ex ante adjustment could be made by replacing the safe interest rate with the interest rate that would apply to a loan with the same default risk that was not accompanied by any financial services. In this example, the spread method would use the 10.25 percent risk-adjusted interest rate instead of the 5 percent safe rate and would tax only the excess above that rate, correctly measuring the financial services.
Although such an approach is theoretically possible,17 it is not practical because risk cannot be accurately measured, as Bradford (1996a, 447) emphasizes. The probability of default is difficult to determine; moreover, if the bank and the borrower are risk averse, a further adjustment must be made to determine the appropriate interest rate.
As Fixler, Reinsdorf, and Smith (2003) explain, the U.S. national income accounts employ the spread method, using a safe interest rate linked to the average rates banks earn on Treasury and federal agency securities. They allude to the possibility of using a risk-adjusted rate as the reference rate, but note that the use of a safe rate accords with international practice, as set forth in the System of National Accounts. If the national income accounts have not yet identified a risk-adjusted rate to apply, purely for statistical purposes, to the financial sector as a whole, it is unlikely that a tax system can identify risk-adjusted rates for each of the millions of financial transactions in the economy, particularly rates sufficiently reliable to use in tax computation.
A more workable version of the spread method adjusts for risk, ex post. This version still uses the safe interest rate, but in a different way. The bank is not taxed on the difference between the stated interest rate and the safe rate, but instead on the difference between the actual payments it receives and those it would have received if it had earned the safe rate. In this example, if the borrower repays the loan with its 11.3 percent interest rate, which has a twenty in twenty-one probability of occurring, the bank is taxed on $63 because it has earned a surplus of that amount over the safe rate. So far, this is the same as the basic spread method, with no risk adjustment. But if the borrower defaults, which has a one in twenty-one probability of occurring, the bank is taxed on negative $1,050 because it receives nothing, which is $1,050 less than it would have received if it had earned a 5 percent return on the $1,000 loan.
At first glance, this ex post spread method appears to completely fail at measuring the financial services provided to the borrower. The borrower receives $10 of services in all cases, yet the method never taxes the bank on $10. There is a twenty in twenty-one probability that the bank will be taxed on $63, which is much higher than the value of the services actually provided, and a one in twenty-one probability that it will be taxed on negative $1,050, which is utterly removed from the correct value.
In reality, though, the method does work because the expected present value of the taxable amount equals the value of the financial services. (Note that [20/21]*[63] + [1/21]*[–1,050] = 10.) The method adjusts for the risk of default, not in advance but rather by making an allowance for it if and when default occurs. This has the tremendous advantage that the tax system need not know anything about the probability of default, as it would have had to do in order to make the ex ante adjustment rejected above. Because the ex post method makes the adjustment if and when default occurs, the probability of the adjustment automatically equals the probability of default, whatever that may be.
Although we have assumed risk neutrality in the above discussion, the ex post spread method also works under risk aversion. With risk aversion, the compensation that the bank demands for default depends not only on the probability of default but also on how averse the bank's stockholders are to the potential losses. For example, if the outcome in which default occurs is one in which stockholders are already doing poorly, making them particularly reluctant to bear further loss, the compensation they demand for default risk will be greater than the expected value of the loss. Although an ex ante adjustment would have to estimate this effect to compute the proper interest rate, the ex post spread method labors under no such difficulty. Because the ex post method provides a deduction for default loss if and when the loss occurs, the deduction has the same greater value to stockholders as the loss, whatever that value might be. This illustrates yet again the point discussed in the "Zero Revenue from Taxation of Risky Returns" box in chapter 3 (pages 51–52), namely, that the taxation of risky returns has zero effect on the government and on investors.
In all of these cases, then, the present value of what is taxed under the ex post spread method equals the value of the financial services. This equivalence of present value is sufficient for tax purposes, as it ensures that all parties face the correct incentives when they make their decisions. Of course, equivalence in present value would not be suitable for national income accounting, which seeks to measure the financial services actually provided in each year.
Assessing the Ex Post Spread Method. Although the ex post spread method is a viable approach, it becomes more complex when it is applied to financial transactions other than simple loans. In many transactions, there are numerous cash flows in each direction, and it is necessary to track what each party receives, relative to a safe return on all of the past cash flows.
The risk adjustment represents a departure from the real-based nature of the X tax. The bank is taxed on the $10 of services it provides to the borrower, plus a random term that has a twenty in twenty-one probability of equaling $53 and a one in twenty-one probability of equaling negative $1,060. The second term has zero expected present value, but its ex post value depends on whether the borrower pays or defaults, which reflects a financial transaction that a fully real-based tax would ignore.
This departure from the real-based system is not, however, a weakness of the method. Indeed, simplicity can be promoted by taking a further step in this direction, as is done by the real-plus-financial (R+F) cash-flow method.
We first discuss a pure R+F cash-flow method and then consider an important revision required to make the method practical. This method has also been discussed by Hoffman, Poddar, and Whalley (1987), Merrill and Edwards (1996), Merrill (1997), and Poddar and English (1997).
Pure R+F Cash-Flow Method. As we emphasized in chapter 2, the X tax is relatively simple because it is a real-based tax that ignores financial flows. Unlike the PET, the X tax does not attempt to track cash flows to their ultimate destination among households. In the context of financial services, however, the real-based nature of the tax creates problems.
The R+F cash-flow method that we now consider departs from the X tax's real basis by taxing financial intermediaries' transactions on a real-plus-financial basis. As we discussed in chapter 2, financial transactions have zero expected present value, so taxing both real and financial cash flows results in the same expected-present-value tax liability as taxing the real alone, assuming that tax rates remain constant. Given that the expected-present-value tax liability ends up being the same, it is ordinarily simpler to tax only the real flows and to ignore the financial transactions. But in the present context, where the challenge is to distinguish real and financial flows, it is simpler to tax all the flows, thereby avoiding the need to disentangle the two categories.
In the above example, under the R+F cash-flow method, the bank initially deducts the $1,000 that it lends to the borrower and is then taxed on any payments received. Under the above assumptions about default risk, there is a twenty in twenty-one probability that the bank is taxed on a $1,113 repayment and a one in twenty-one probability that it is taxed on nothing. The expected value of this taxable amount, minus the initial $1,000 deduction (compounded forward with 5 percent interest), is equal to the $10 cost of the financial services provided to the borrower.
The R+F cash-flow method is similar in many respects to the ex post spread method. Under each method, the ex post tax payment ultimately made by the bank depends on whether default ends up occurring, which has nothing to do with the cost of the financial services. In each case, though, the expected present value of the uncertain tax payment equals the cost of the services. By making the tax payment depend on the outcome of a financial transaction, each method departs from the real-based nature of the X tax.
The R+F cash-flow method is simpler to administer than the ex post spread method because there is no need to track past flows and apply the safe interest rate to them. In the example, the ex post spread method requires that the bank's initial outlay of $1,000 to the borrower be tracked, so when the bank subsequently receives payments, it is taxed only on the difference between its receipts and a safe return, computed from the time of the outlay to the time of the subsequent payment, on the $1,000. The R+F cash-flow method is simpler. The $1,000 payment is deducted when it is made and thereafter ignored, as the bank is taxed in full on subsequent receipts from the borrower. The simplicity advantage is magnified in situations involving multiple payments in both directions. Bradford (1996a, 460) aptly states that "cash-flow approaches promise relatively simple solutions."
One trade-off for this simplicity is a loss of intuitive appeal. (Another, discussed below, is the need for transition arrangements). The R+F cash-flow method is highly counterintuitive in many contexts, including this one. The bank deducts the $1,000 it lends to the borrower, even though such lending is in no sense an expense or loss, and is then taxed on the borrower's subsequent payment, including the repayment of principal. The method also applies to the checking account, with the bank paying tax on deposits into the account and deducting withdrawals, including interest, from the account. The inclusions and deductions not only have no visible relation to the cost of financial services (as was also true under the ex post spread method) but also have little visible relation to income and loss. Discussing R+F cash-flow methods, Shoven (1996, 462) notes that "their appearance can be quite different from their actual economic incidence, and this can make them less politically viable."
Nevertheless, we recommend the use of the R+F cash-flow method. However counterintuitive the bank's calculations may be, they are simple to perform, and they correctly measure the expected present value of the financial services.
But one challenge yet remains. Under the normal working of an X tax, firms are taxed on sales to both households and firms, and firms deduct their purchases from other firms. If the R+F cash-flow method is applied in accord with that principle, then, if the bank's customer is a business firm, it must perform the mirror image of the bank's computations in order to deduct (in expected present value) its purchase of financial services. For example, the firm must deduct each bank deposit it makes and pay tax on each withdrawal, including each check written.
Such a policy is obviously unacceptable. The owner of a local hardware store would surely not comprehend the logic of paying tax on each check the store writes and would not be mollified by the assertion that doing so helps tax the expected present value of financial services received on the checking account. As we noted earlier in this chapter, an acceptable method cannot require ordinary businesses to engage in counterintuitive tax computations. Moreover, the transition arrangements, discussed below, would be impossible to implement if they had to apply to all firms, rather than merely to financial institutions.
Some proposals that apply the R+F cash-flow method, including Poddar and English (1997), address this issue by requiring the bank to allocate its tax liability among its customers and to inform each business customer of the amount it can deduct. As Merrill (1997, 35–36) notes, such allocations would be complex.
Limiting the Method to Transactions with Households. Fortunately, this problem can be solved by modifying the R+F cash-flow method to have the tax system ignore transactions between financial institutions and business firms, an option discussed by Bradford (1996a, 448) and Merrill (1997, 36). Under this approach, financial institutions are not taxed on their sales of financial services to firms, and firms do not deduct their purchase of such services. Financial institutions are taxed, using the R+F cash-flow method, only on their transactions with households, which do not claim offsetting deductions. Under this modification, only the financial institutions need to use the counterintuitive R+F cash-flow method. (In chapter 7, we will also recommend the application of the R+F method to transactions between U.S. firms and their foreign affiliates, on the premise that it is acceptable to require counterintuitive tax computations by multinational firms.)
Recall that the X tax, like the VAT, imposes no net tax on transactions between firms, as the taxation of the seller is offset by a deduction for the buyer. The retail sales tax reaches the same net result by simply ignoring both ends of the transaction and taxing only sales to consumers. Our proposal effectively applies the sales-tax approach to this set of transactions. Although enforcement for a high-rate tax is normally improved by having an offsetting inclusion and deduction for sales between firms, it is preferable to ignore the transaction if doing so allows ordinary businesses to escape counterintuitive tax computations. This proposal illustrates the principle we mentioned in chapter 2, that the X tax can mix and match different methods in different contexts to achieve the best administrative results.
Bradford (1996a) and Merrill (1997) each note that there would be some complexities associated with requiring financial institutions to distinguish between transactions with firms and those with households. (Retail sales taxes face this same challenge with respect to a wide range of transactions.) But the complications are surely manageable. Customers should be presumed to be households unless they demonstrate business status and file business tax returns of their own.
The success of this proposal relies on financial institutions charging different prices, either in terms of interest rates or fees, to business and household customers. On two otherwise identical loans, the bank must charge a tax-inclusive price to a household borrower and a tax-free price to a business borrower. Recall that the X tax or VAT normally does not require such differential pricing; households and business firms buying computers pay the same tax-inclusive price, with the firms, but not the households, claiming offsetting deductions on their own tax returns. In contrast, the retail sales tax, which we are imitating in this context, requires such differential pricing, as firms charge tax to their household customers, but not their business customers. If financial institutions are constrained to charge uniform prices, then the method will not work. Any regulatory barriers to differential pricing should therefore be removed as part of the tax reform.
Unified Treatment for Financial Firms. One advantage of the R+F cashflow method is that it applies uniformly to almost all financial institutions. There is no need for separate rules for banks, insurance companies, and other institutions. Although it is necessary to distinguish financial institutions from nonfinancial firms, it is generally not necessary to distinguish among financial institutions. Almost all financial institutions are subject to the same rules, paying tax on cash inflows and deducting cash outflows. As discussed above, banks and thrifts pay tax on deposits received, and loan proceeds collected, from households and deduct deposit withdrawals by, and loans made to, households. Similarly, property and casualty insurance companies pay tax on premiums paid by households and deduct benefit payments made to households; there are no loss reserves. The R+F cash-flow method's application to insurance companies is more intuitive, and more similar to current law, than its application to most other financial firms.
As discussed below, however, there may be some special cases in which the method cannot be applied, such as to some transactions of securities dealers. Special rules will be required for those cases.
Coverage. With respect to the coverage of the R+F cash-flow method, two dimensions must be addressed. First, it is necessary to determine which firms are financial institutions and therefore subject to the method on some or all of their transactions. Second, for those institutions, it is necessary to determine which of their transactions are subject to the method.
The first question is relatively easy to address. A simple rule is that the R+F cash-flow method should apply to any firm that is regulated as a financial institution by any federal or state agency under any statute. If a firm is not regulated as a financial institution, it is unlikely that it provides a significant amount of financial services.
The second question is more difficult, as it may often not be possible to precisely identify the transactions that include financial services for which explicit fees are not charged. It is probably better, when in doubt, to err on the side of subjecting transactions to the R+F cash-flow method. If the transactions actually include financial services, the failure to apply the R+F method allows financial services to escape tax. On the other hand, if the transactions actually do not include financial services, the application of the R+F method does not lead to any mismeasurement of the tax base, because the R+F method imposes zero tax in expected present value on transactions that do not include financial services. At worst, the unnecessary application of the method could cause additional complexity, but even that need not occur. Given that some of a firm's transactions are subject to the R+F method, complexity has no necessary relationship to the volume of the transactions subject to the method. Instead, complexity depends on how readily the firm can determine which transactions are subject to the method.
Accordingly, if an institution is subject to the R+F cash-flow method, the method should apply to all of its transactions, except those that are specifically removed by regulation. Transactions should be removed only if they clearly involve no significant provision of financial services, such as the issuance of securities in organized markets, or if any services are clearly financed in full through explicit fees. Bright lines should be drawn to minimize complexity. If one set of transactions includes real services disguised as financial flows while another set of transactions does not, and no simple rule can be written to distinguish between them, then both sets of transactions should be subject to the R+F cash-flow method.
Transition. As discussed above, the R+F cash-flow method relies on the expected present value of the cash flows being equivalent to the cost of the financial services. Because this equivalence holds only over the entire life of the transaction, the method faces significant transition issues, as Bradford (1996a, 443) notes.
In the above example, suppose that the $1,000 loan was made before the X tax's effective date, but the borrower's repayment occurs after that date. Then, the bank is taxed on the borrower's potential repayment, which has an ex ante market value of $1,010, far more than the $10 cost of the financial services. The problem is the missing $1,000 deduction for the initial loan. For a checking account, the opposite problem arises. If an account is withdrawn after the reform, the bank deducts the withdrawn amount, although it was never taxed on the initial deposit.
The problem would be solved if the bank could include the market value of its outstanding deposits and deduct the market value of its outstanding loans on the effective date. Although market value is difficult to measure, a ready proxy is provided by the tax basis that banks have in their accounts and loans. If subsequent rate changes are large, they should be handled in a similar manner.
Many transactions in the nonfinancial sector, such as sales on credit, involve a bundling of real and financial transactions. Suppose, for example, that a firm sells apples to a consumer at a stated price of $100 and gives the consumer one year to pay the $100 without any stated interest. Also assume that the consumer makes the $100 payment on schedule. To tax consumer purchases, the X tax system should ideally impose tax at the sale date on the value of the apples. Unfortunately, that value is not readily determinable, although it is clear that it must be less than $100. In economic terms, the $100 payment includes an interest payment to compensate for the deferral of payment, even though the firm has chosen not to label any of the payment as interest. If tax is to be imposed at the sale date, the tax should not apply to the interest component of the $100 payment, as the X tax ignores interest income and expense. The tax should apply only to the underlying value of the apples.
As Bradford (1996a) notes, however, a simple solution is available. It once again involves the R+F cash-flow method, which, in this context, simply means using a cash method of tax accounting. The cash method taxes all of the cash flows arising from the sale of the apples when the cash flows occur; in this example, that involves imposing no tax on the sale date and instead taxing the firm on the $100 payment when it is received. The cash method works because the present value of the deferred $100 payment must be equal to the value of the apples on the sale date. The cash method continues to work in the presence of default risk, just as it did for the bank loan discussed above, because the expected present value of the uncertain future payment equals the value of the apples on the sale date. If there is a risk of default, the firm increases the stated price to compensate for the possibility that no payment will be received. The cash method then taxes the increased payment if the increased payment is received and taxes nothing if nothing is received; the expected present value of what is taxed then necessarily equals the value of the apples on the sale date.
So, the cash method imposes no tax and offers no deduction when goods are exchanged for a receivable or payable. Instead, inclusions or deductions are made with respect to all future cash generated by the receivable or payable. Under the cash method, it is not necessary to specify a time at which the sale is deemed to have occurred; only the dates of receipts and payments need be determined.
For example, a firm that sells on credit is not taxed at the time of sale. Instead, it is taxed when it collects on the resulting account receivable, regardless of whether the collection is labeled as principal or interest. If it does not collect, due to the buyer's default, no tax is paid. It is also taxed if it sells the receivable.
Conversely, a firm that buys on credit claims no deduction at the time of sale. Instead, it deducts any payment on the resulting account payable, regardless of whether the payment is labeled as principal or interest. A buying firm also deducts the cost of paying cash to a third party to assume an account payable.
Strictly speaking, the expected present values of the tax payments are correct only if tax is paid at the same instant that the payment is made or received. Weisbach (2000, 637–39) observes that the cash method works less well if firms' accounting periods are long relative to the durations of the financial items, because tax payments then do not coincide with the recognition of the cash flow. Under current law, C corporations are required to pay estimated tax on a quarterly basis; the X tax should also require quarterly tax payment by firms.
The cash method also provides proper treatment, in expected present value, for contingent payments, deposits, defaults, and returns. The method also easily handles leases, including leases with options to purchase. The method simply includes the cash that changes hands, all of which is payment for either the use of the leased item or the possibility of purchase. Note that the latter payments are treated correctly in expected present value, regardless of whether any purchase ultimately occurs.
The cash method is similar in some respects, and different in others, from the methods typically used in VAT systems. As Weisbach (2000, 628, 640) and Tait (1988, 373–90) discuss, European VATs typically tax the purchase price of the good or service at the time that the good or service is deemed to be supplied. Tait notes that the time of supply can be defined as the time when the invoice is issued, when the goods are made available or the services rendered, or the time of payment, if earlier. The measurement of the purchase price can also be difficult. Tait explains that this price is normally net of discounts, that contingent payments are generally taxed only when they are determined, that customer deposits are usually not taxed until the purchase is finalized, and that tax is rebated if items are returned for credit. For an installment sale, the stated principal may be taxed at the time of sale, with no tax on the stated interest. The cash method offers a simple unified approach that can handle all these situations.
In some circumstances, application of the R+F cash-flow method is difficult, potentially warranting the use of alternative methods. Problems would occur, for example, if the cash flows arising from a sale stretch out over an extended period.
In the tax accounting context, suppose that the firm is paid for the apples with a share of stock. The mechanical application of the cash method would require that the firm pay tax on all dividends received while it holds the share and on the gross proceeds from any subsequent sale of the share. In principle, the cash method still works in this context; because the firm accepted the share as payment for the apples, the expected present value of the dividends and sale proceeds must be equal to the value of the apples on the sale date. If the firm were to hold the share for an extended period, though, the cash method would prove cumbersome, as it would require the imposition of tax on dividends received years after the sale of the apples. It would be preferable to instead tax the receipt of the share by treating it as a "cash" receipt in the amount of the share's market value. Of course, few consumers make payments of this kind.
Problems of this kind could also arise in the context of financial institutions. Consider, for example, a dealer who buys and sells securities and receives compensation for his financial services through the bid-ask spread rather than through explicit fees. If the dealer purchases securities from a household, the mechanical application of the R+F cash-flow method calls for the dealer to deduct the purchase price and to pay tax on the subsequent cash flows from the sale of the security. If the security is sold to a business, though, the cash flows are improperly measured, because the sale proceeds would include the implicit value of services provided to the business purchaser, services on which the dealer is not supposed to be taxed. Here, too, it may be necessary to impute a market value to the security at the time it is purchased from the household or to adopt some other special rule.
Weisbach (2008, 85–86) also observes that the R+F cash-flow method does not work for barter transactions, in which goods are exchanged for goods or assets. Here, too, it may be necessary to impute the market value of the goods.
The advantage of the real-based approach that underlies the X tax is that it allows the tax system to ignore financial flows. Its disadvantage is that it requires the tax system to distinguish between real and financial flows. The best approach is to use the real-based approach when it is easy to distinguish the two and to use a real-plus-financial cash-flow method when it is difficult to distinguish between them. In that spirit, we recommend that a real-plus-financial cash-flow method apply to those transactions between financial institutions and households in which real services may be disguised as financial flows.