Appendix I. Measurement and Accounting Issues

George Mackenzie, and Peter Stella
Published Date:
October 1996
  • ShareShare
Show Summary Details

Various problems that arise in defining QFA, and in measuring the quasi-fiscal component of the operations of central banks and other PFIs, were briefly discussed in the text, as were the difficulties that can ari0se in consolidating the financial operations of central banks with those of the NFPS. This appendix will address these issues at greater length.

Definitional Issues

A good starting point for the discussion is a modified version of the narrow definition of QFA set out in Section II: an operation or measure carried out by a central bank or other PFI with an effect that can, in principle, be duplicated by budgetary measures in the form of an explicit tax, subsidy, or direct expenditure and that has or may have an impact on the financial operations of the central bank or other PFIs.

This definition merits a number of observations:

  • It excludes certain activities that in some countries have entailed large losses for the central bank. One topical and important example is the losses that have been incurred by policies to sterilize capital inflows; another is the impact on a central bank’s income statement generated by a restrictive open market operation.35 There is a sense in which these losses—or any central bank loss—are fiscal, inasmuch as they ultimately have to be covered by the government. Although they do not have to entail a tax or subsidy element, they are commonly referred to as QFAs.
  • Consistent application of the definition may require that certain activities normally considered monetary in character be recognized as having a fiscal element. For example, most central banks impose a reserve requirement, and many of these do not pay a market-related rate of interest on their reserves. If a reserve requirement results in banks holding a greater share of their assets in reserves than they would in the absence of that requirement, it imposes a tax, as discussed in Section II.36 Yet this practice may be seen by some as an integral part of monetary policy.
  • The definition would also exclude some government regulations and other acts of public policy that have little or nothing to do with either the budget or monetary policy but that can have effects akin to those of taxes and subsidies. For example, a regulation restricting the use of property can inflict a capital loss on the owner, and in this respect it is similar to a tax on wealth. The definition excludes these practices from the ambit of QFA, because they do not directly affect the financial operations of PFIs.37
  • It would not include tax expenditures, which are already covered by the budget and in principle are subject to budgetary scrutiny. This exclusion does not mean that a government should not be concerned about the fiscal cost of special deductions from taxable income and the like. These are clearly important, but they are rather different from the quasi-fiscal activity pursued by PFIs. Similarly, the definition excludes the estimated cost of loan guarantee programs created by the government and other programs entailing contingent liabilities. As the paper stresses, these operations are potentially very costly, whether conducted by the government or by a PFI. See Towe (1993) for further discussion.
  • The definition would exclude the inflation tax, in view of the following considerations:
    • —The inflation tax is not really an “operation” or a measure, but the end result of a particular combination of monetary and fiscal policies. The term “tax” is, consequently, somewhat misleading.
    • —The replacement of an explicit tax on money balances by the inflation tax entails a wholesale change in the macroeconomic environment (that is, an increase in the rate of inflation). Replacing an export tax by an MER practice has, by contrast, no such effect.
    • —The rate of the inflation tax will not typically equal any given rate of an explicit tax on money balances.

Another implication of the definition is that its strict application would exclude certain types of operations that entail the creation of taxes and subsidies through the financial system when these have no direct impact on the net income of the central bank or other PFIs. As discussed in Section II, the central bank may establish regulations setting maximum lending rates and deposit rates for commercial banks, both privately and publicly owned. Artificially low deposit rates are like a tax on savers, and low interest rates are a subsidy to borrowers. Yet the net income of PFIs may be entirely unaffected by the regulations (that is, if there are no publicly owned commercial banks). These kinds of regulations bear a certain resemblance to employer-mandated benefit plans. Nonetheless, these operations do affect the financial operations of the government, as previously discussed, because they artificially lower the cost of borrowing.

Since the spirit of these regulations is the same as preferential interest rates and other QFAs, and since they impose a tax that ultimately benefits the government, the definition could be expanded to include them. The modified definition would be as follows: an operation or measure carried out by a central bank or other PFI with an effect that can, in principle, be duplicated by budgetary measures in the form of an explicit tax, subsidy, or direct expenditure and that has or may have an impact on the financial operations of the central bank, other PFIs, or government.

In practice, QFA is defined even more broadly, since it can refer to any operation that has a significant impact on the net income of a PFI. That said, there is some merit to distinguishing between such practices as MERs or subsidized interest rates, which clearly entail taxes and subsidies, and practices such as open market operations, which have fiscal implications but cannot be so obviously duplicated by explicitly budgetary operations. The paper has relied on a definition of QFAs broad enough to encompass these kinds of operations, while paying special attention to those operations of a more obviously fiscal character.

Problems Resulting from the Use of Cash Versus Accrual Accounting

Most Fund member countries measure the financial operations of the NFPS or central government on what might be called a modified cash basis—taxes and other revenues are recorded as collected, and expenditures when they are paid, with interest recorded when due, and sometimes with an adjustment for changes in arrears. Financial institutions, and central banks in particular, will typically use accruals-based accounting. To take a few examples: interest on loans is recorded when it is due, not when it is paid; provisions for doubtful loans, loss write-offs, or loss reserves will be made that have no counterpart in a cash transaction; depreciation of physical assets is recorded as an expense; expenditure on capital assets is recorded in the capital account and does not affect the net income position. For all these reasons, it cannot be assumed that the cash transfer made by the central bank to the budget in a given period will equal its cash income for that period, measured in the same way as the operations of the NFPS, even if it transfers 100 percent of its accrued income to the central government.

To derive the enlarged public sector balance described earlier—to amalgamate central bank or other PFI income (losses) with the balance of the NFPS in a consistent way—requires that some fairly major adjustments be made. However, since the operations of the NFPS or the central government are not measured entirely on a cash basis, neither should the operations of the PFIs. For example, if interest paid to the central bank by the central government is measured on an accrual basis the bank’s receipts should be measured in the same way to avoid inconsistency. There is also a potential problem with a cash-based measure of central bank income: it has happened that borrowers in arrears on their interest obligations to public sector banks have obtained fresh loans to pay off their arrears. This kind of operation can be used to reduce artificially the overall public sector deficit. To avoid this kind of manipulation, appropriate bank supervisory regulations prohibiting such recapitalization of interest should be put in place.

The accounting treatment of net lending by the central bank for policy purposes deserves particular attention. Such operations, if conducted through the budget, are recorded as above-the-line transactions in the framework of the IMF’s government finance statistics (International Monetary Fund, 1986). Hence, they increase the deficit. Typically, they would not be treated this way in the accounts of the central bank.

Particular Problems Resulting from Foreign Exchange Transactions and MER Systems

This section begins with a brief discussion of accounting procedures for spot foreign exchange transactions without the added complication of the quasi-fiscal taxes and subsidies entailed by MERs.38 It then takes up the special problems posed by MERs.

In its role as custodian of a country’s foreign exchange reserves, a central bank is constantly accruing or realizing losses or gains in its foreign exchange transactions. Considering first losses or gains that are not realized—those that result from a change in the value of the stock of foreign exchange—the standard treatment, whatever the precise rule used to value foreign exchange, is to make a counterpart entry in the revaluation account. If central banks make a daily valuation of their reserves based on the average daily rate, then the revaluation account is also changed on a daily basis. These accrued gains or losses, however, are not reflected in the central bank’s income and loss statement.39 This treatment of accrued losses or gains is consistent with the standard presentation of the financial operations of the NFPS. The same conclusion holds true when valuations are less frequent, provided that a counterpart entry is made in the revaluation account.

This use of the revaluation account makes no distinction between nominal and real gains and losses. Thus, for example, the central bank’s foreign exchange reserves can actually increase or decrease in real terms without there being any effect on the income and loss statement. Such accrued losses or gains, however, do affect the strength of a country’s reserve position and cannot be entirely ignored.

In general, recommended accounting practice calls for only realized gains on foreign exchange transactions to be brought to the profit and loss account. A number of central banks (perhaps most prominently the U.S. Federal Reserve System), however, include both realized and unrealized gains and losses in the profit and loss statement; that is, they include the valuation change on the stock of net foreign assets (NFA). In such cases it is clear that, in order to ensure consistency with government accounts, the central bank operating result would need to be adjusted before amalgamating it with NFPS operations.

Given the way realized gains are frequently measured, their inclusion could be problematic. To take a concrete example, suppose a central bank had acquired its target level of foreign exchange reserves at an average rate of LCU 1 per U.S. dollar, and that the exchange rate had subsequently been devalued to LCU 2 per dollar. At the time of the devaluation, a revaluation gain of LCU 1 million for each $1 million of reserves would have been recorded. Without any sales of foreign exchange, this gain would not have affected the profit and loss account.

If the bank now sells $1 million at LCU 2 per U.S. dollar, the monetary base and NFA each decline by LCU 2 million. If it subsequently repurchases the foreign exchange with monetary base at the same rate—LCU 2 per dollar—the earlier transaction is reversed. In determining profits and losses for the year, however, the bank’s accountant would debit the revaluation account by the amount of profit deemed to be realized from the sale of foreign exchange. If the foreign exchange sold was valued at LCU 1 per dollar, the sale is recorded as generating a profit of LCU 1 million, which is added to the profit and loss account and subtracted from the revaluation account.

The economic justification for distinguishing this as realized “profit” is rather dubious, however. In the example given, the balance sheet would have been identical had the central bank undertaken no intervention (presuming the final exchange rate was unaffected by the intervention). The only difference would be that in the latter case the accountant would not have debited the revaluation account and added the amount to the profit and loss statement, since no transactions had taken place.

An analogy could be made with the use of first in, first out (FIFO) accounting by any enterprise with inventories in an inflationary period. Historic cost pricing means that accounting profits are artificially overstated; the cost to the enterprise of replacing its inventories has risen, and once it has replaced them it is in the same position as before with the exception that the value of its inventory has risen. The profit is “locked up” in the inventory, however, and the enterprise does not realize this gain unless it reduces its stock of inventory. The same is true of the central bank. If it maintains an unchanged target for NFA, it will not effectively realize a gain on its foreign exchange holdings despite gross sales and purchases in the market.

We now consider the valuation and classification problems posed by MERs. Two cases should be distinguished: when the central bank’s NFA do not change, and when they do. As a simple example, take a three-rate MER system in which the officially designated central rate is LCU 2 per dollar, with a special appreciated rate of LCU 1 per dollar applying to certain exports, and a depreciated rate of LCU 3 per dollar applying to certain imports.

In the first case, where net sales at the central rate are zero, the sale of $1,000 million to importers at the special rate and the purchase of $1,000 million from exporters at their special rate entail no change in NFA, a reduction in the monetary base (or negative monetary impact) of LCU 2,000 million, and an increase in OIN (other items net) of LCU 2,000 million, reflecting the central bank’s profits from the operation (Table 2). These calculations are not affected by the choice of the central rate, even if that choice is to some extent arbitrary, because there is no change in NFA.

Table 2.Impact of an MER System on the Central Bank’s Balance Sheet: With No Change in Reserves
A. Foreign Exchange Transactions and Their Monetary Impact
Exchange Rates (LCUs per dollar)
123All Rates
Net sales (net purchases –)
(in millions of dollars)–1,0001,000
Monetary impact
(in millions of LCUs)1,000–3,000–2,000
B. Impact on the Central Bank’s Balance Sheet (in millions of LCUs)
Monetary impact–2,000
Change in NFA
Change OIN
(central bank profits)2,000
Note: LCUs, local currency units; NFA, net foreign assets; OIN, other items net.
Note: LCUs, local currency units; NFA, net foreign assets; OIN, other items net.

What would be affected by the choice of the central rate, however, is the classification of the quasi-taxes or subsidies entailed by the MER system. If, in the example just presented, the officially designated central exchange rate applies to most transactions, then this may not be an issue. However, the structure of a multiple rate system might be such that there were at least two plausible candidates for the central rate. In this case, any measure of the tax or subsidy equivalent of special exchange rates will inevitably be arbitrary.

When the level of international reserves changes, the choice of the central exchange rate does matter for the calculation of central bank income. To return to the earlier example, now augmented by a fourth exchange rate of LCU 2.5 to the dollar, let us suppose that net sales of foreign exchange at the rates of LCU 2 and LCU 2.5 are zero, but that sales of foreign exchange to importers at the rate of LCU 3 to the dollar are now $500 million, not $1,000 million.

The monetary base is reduced by LCU 500 million, but the measured impact on OIN clearly varies depending on which exchange rate is used to value the change in reserves (Table 3). This result is a simple illustration of the inventory valuation problem, but it does illustrate how different accounting conventions will affect the measure of central bank income.

Table 3.Impact of an MER System on the Central Bank's Balance Sheet: With Change in Reserves
A. Foreign Exchange Transactions and Their Monetary Impact
Exchange Rates (LCUs per dollar)
122.53All Rates
Net sales (net purchases –)
(in millions of dollars)–1,000500–500
Monetary impact
(in millions of LCUs)1,000–1,500–500
Note: OIN, other items net.
B. Impact on the Central Bank’s Balance Sheet (in millions of LCUs)
With Exchange Rate of LCU 2 = $lWith Exchange Rate of LCU 2.5 = $l
Monetary impact–500–500
Change in NFA1,0001,250
Change in OIN (central bank profits)1,5001,750
Note: OIN, other items net.
Note: OIN, other items net.

The conventional approach to the valuation of reserve gains and losses—the use of the rate at which most official transactions take place—becomes even more problematic when none of the official exchange rates is close to the free-market rate. In such a case, the use of any official exchange rate for valuation purposes can be quite misleading. To take an extreme example, suppose that the free-market rate in the example above is LCU 10 to the dollar. If this rate, rather than an official rate, is used to value the reserve gain, the calculated impact on central bank income is substantially different. Even when there is no change in NFA, the use of official exchange rates in such a situation to measure the relative size of the implicit taxes and subsidies created by the exchange system will be seriously misleading.

Reserve Requirements

Section III discussed the role of reserve requirements as quasi-fiscal taxes. When banks are obliged to hold reserves and are paid a lower rate of interest than they would earn on an alternative and equally attractive investment, they are in effect being subject to a tax. If banks’ reserves are lent to the government, and if this lending is a substitute for sales of government securities to the private sector, then the government is in effect given an interest subsidy that equals the difference between the rate of interest on government bonds (ib) and the rate of interest paid to banks on their reserves (ir) times the stock of reserves (R). The amount of the subsidy (S) is thus given by

S = (ib-ir)R.

If reserves are a certain fraction (k) of deposits (D), the subsidy may be re-expressed as:

S = (ib - ir)(kD).

This subsidy can affect the financial operations of the government in one or a combination of two ways. If the central bank does not charge interest on its loans to the government, its net income is reduced by the imposition of—or increase in—a reserve requirement, since it may pay interest on the bank reserves. This means that transfers of central bank profits to the government will be lower than otherwise.40 However, this negative impact on the government’s accounts will be more than offset by lower explicit interest payments by the government, since ib exceeds ir. Although government borrowing is subsidized, the explicit rate of interest on government securities is not reduced; instead, the amount of borrowing by this means is reduced.

If the central bank does charge a market rate of interest on its loans to the government, the government’s interest expenditures will not be affected. The central bank’s net income will be affected, however, and the increase in its transfer to the government will reduce the budget deficit. Despite the fact that interest expenditure is not affected, the imposition of a reserve requirement where reserves are compensated at a rate below the market effectively lowers the government’s cost of borrowing.

The reserves tax does not have to be used to subsidize lending to the government; it can be used to finance any expenditure. Nonetheless, the use of artificially high reserve requirements to channel credit to the government is certainly not uncommon.

The quasi-fiscal tax on reserves can be made transparent by including it—for analytical purposes at least—in government tax revenue, which will then increase by the value S. If the central bank has not been charging the government interest on the loans that are financed by the increase in reserves, this should be offset by an increase in interest payments of the same amount. When the central bank has been charging interest, the increase in tax revenue should be offset by a decline in central bank income —if a marginal rate of transfer of central bank profits of 100 percent is assumed, then the government’s property income is reduced by the amount of the tax.

This discussion has focused on how the tax entailed by reserve requirements is determined, and on the way it shows up in the public sector’s financial accounts. The incidence of the tax is another matter. To the extent that reserve requirements lower deposit rates, for example, then depositors bear part of the cost; higher lending rates mean that borrowers pay part of it as well. These issues have been discussed at some length in Molho (1992).

Contingent Liabilities

The problems posed for the contingent liabilities entailed by a central bank’s quasi-fiscal operations are essentially the same as those posed by similar operations of the central government. Both central banks and central governments can guarantee loans, for example. In both cases the size of the operations can be very large, but it will have no impact on a cash-based measure of the financial operations of the government until money is actually paid out for a loss. The inadequacies of a cash measure of the public sector’s operations, when contingent liabilities are important, have been the subject of some commentary (see Towe, 1993, for example). At a minimum, cash-based measures of the deficit should be supplemented with a measure of the outstanding value of contingent liabilities.

Substantial valuation problems may arise with contingent liabilities; these problems are not unique to the operations of central banks and other financial public institutions. To take the case of loan guarantees, should one seek data on the total stock of loans that are guaranteed or just that part of the stock that is deemed to be at risk of non performing? Often the information that would permit the calculation of the expected value of realized losses entailed by a contingent liability program may not be available.41

Towe (1993) has described one method for calculating the capitalized subsidy element created by a loan guarantee. This is the present value of the reduction in interest payments that results from the guarantee and is given by the following formula, where L is the loan principal, iw is the interest rate that would have been obtained without the guarantee, and ig is the guaranteed rate:

Inflation Adjustment

The adjustment of the conventional measure of the financial balance of the NFPS to take account of the impact of inflation on the real value of net public debt is a controversial issue.42 Nevertheless, inflation-adjusted deficits can play a useful complementary analytical role to traditional indicators of the fiscal stance. The basic rationale for the adjustment is that the inflationary component of the government’s interest expenditure is really a form of amortization. It compensates the holders of public sector debt for the decline in the real value of their assets; therefore it should not be classified as a current expense of the government or as income in the hands of the recipient.43

The income statements of central banks are also affected by inflation, and the issue arises whether the conventional results need to be adjusted for inflation. This section briefly discusses some of the consequences of using the standard inflation adjustment, with the aid of this highly simplified central bank balance sheet:

Net foreign assets (NFA)Monetary base (MB)
Net credit to government (DCG)
Net credit to private sector (DCP)Net worth (NW)

It is assumed, for simplicity, that the bank has no operating costs and no physical assets. Also for simplicity, net worth includes the valuation adjustment for changes in the nominal value of foreign exchange: hence, it reflects not only nominal increases in the bank’s net credit position, other things being equal, but also increases in the nominal (local currency) value of foreign exchange.44

If the real rate of interest, r, on domestic and foreign assets is the same, the real exchange rate is constant, and no interest is paid on the monetary base (MB) then nominal profits can be expressed as

NFA(r+ pw + d) + (DCP + DCG)(r + pd),

where r is the real rate of interest, d is the rate of depreciation of the local currency, and pw and pd are the international and domestic inflation rates.45 This can be reexpressed as

NFA(r + pd) + (DCP + DCG)(r + pd).

Adjusted for inflation in the usual way, central bank income would then become

NFA(r + pd) + (DCP + DCG)(r + pd)-pdNW


r(NFA + DCP + DCG) + pdMB.

But this would include a measure of the inflation tax, pdMB, in the real income of the central bank. Thus, if the “real” component of the central bank’s income were transferred to the government, revenue from the inflation tax would be counted above the line. This suggests that the standard approach to adjusting deficits for inflation is not appropriate in the case of the central bank. Eliminating the inflation tax term from the preceding equation would simply leave:

r(NFA + DCP + DCG).

If, for the sake of argument, it is assumed that the central bank were paying the market rate of interest on the entire monetary base, then the conventional adjustment would result in the following statement of real income:

r(NFA + DCP + DCG- MB).

In this case, however, the central bank would be compensating holders of the MB by paying a market-related rate of interest on its liabilities. The central bank is—by assumption—not exploiting its monopoly over money creation, and it is earning simply the market-related return on its net worth (the expression in parentheses equals NW). In this case, the conventional adjustment would be appropriate.

    Other Resources Citing This Publication