The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
federal Reserve Bankof NeHlbiic Quarterly Review Autum n 1985 Volum e 10 No. 3 1 P ublic and P rivate Debt A c c u m u la tio n : A P erspective 6 Federal Tax R eform and the R egional C h a ra c te r of the M u n icip a l Bond M arket 16 R ecent In s ta b ility in M 1 ’s V elocity 23 The S trong D o lla r and U.S. Inflation 30 Federal D e p o sit In surance and D eposits at Foreign B ranches of U.S. Banks 39 A R M s: T h e ir F inancing Rate and Im pact on H ousing 50 Treasury and Federal R eserve F oreign E xchange O p e ra tio n s The Quarterly Review is published by the Research and Statistics Function of the Federal Reserve Bank of New York. Remarks of E. GERALD CORRIGAN, President of the Bank, on a perspective of public and private debt accumulation begin on page 1. Among the members of the staff who contributed to this issue are ALLEN J. PROCTOR and JULIE N. RAPPAPORT (on federal tax reform and the regional character of the municipal bond market, page 6); LAWRENCE J. RADECKI and JOHN WENNINGER (on recent instability in M1’s velocity, page 16); CHARLES PIGOTT and VINCENT REINHART (on the strong dollar and U.S. inflation, page 23); CHRISTINE M. CUMMING (on federal deposit insurance and deposits at foreign branches of U.S. banks, page 30); and CARL J. PALASH and ROBERT B. STODDARD (on the financing rate and impact on housing of ARMs, page 39). A semiannual report on Treasury and Federal Reserve foreign exchange operations for the period February through July 1985 starts on page 50. Public and Private Debt Accumulation: A Perspective I welcome this opportunity to address the American Bankers Association annual gathering of Chief Financial Officers. In reflecting on possible topics for my remarks, it struck me that this was a good opportunity to raise some questions about an all too well-known four-letter word. That word is “debt”. Specifically, I want to review with you the facts as they pertain to the disturbing rate at which the U.S. economy is accumulating debt; to cite some of the factors which may lie behind the rapid growth in debt; and to make a few suggestions as to ways in which the growth in debt can be— perhaps I should say must be— moderated over time. In each of the past several years total debt in the economy has risen markedly faster than GNP... Over the 1981*85 period, the ratio of debt to GNP will have risen by about 20 basis points to over 1.60—a very large change in a ratio of this nature. By way of a general background, until recently, the growth in total debt in the economy tended to track closely the growth in nominal GNP. To be sure, there were some departures from this pattern for cyclical and other reasons, but the long run parity between the growth in debt and the growth in GNP was strikingly similar. But, beginning in the 1981-82 time frame Remarks of E. Gerald Corrigan, President, Federal Reserve Bank of New York, before the American Bankers Association Chief Financial Officers’ Forum on Wednesday, September 18, 1985. something seems to have happened to that relationship. In each of the past several years debt has risen mark edly faster than GNP. In fact, using 1981 as a base, the cumulative gap between the growth in debt and the growth in GNP is fifteen percentage points. Stated somewhat differently, over the 1981-85 period, the ratio of debt to GNP will have risen by about 20 basis points to over 1.60— a very large change in a ratio of this nature. A straight extrapolation of this recent trend over the next decade would suggest that by 1995 we would have about $2.25 in debt for every dollar of GNP. In a proximate sense, it is widely recognized that the major factor contributing to the rise in total debt in recent years has been the string of massive Federal budget deficits which have been chalked up in the decade of the 80s. While that is certainly true, the rate at which debt is being accumulated in the private sector is also cause for concern. Let me cite a few statistics that seem particularly telling. In the Federal sector, commentary about $200 billion deficits is now so commonplace that we may tend to lose sight of the financial implications of those mega deficits. For example: • This year, interest costs of servicing the burgeoning Federal debt will total about $130 billion. That will be roughly equal to total personal income tax col lections from every taxpayer west of the Mississippi River. At the same time, more than $20 billion of those interest payments will go to foreign holders FRBNY Quarterly Review/Autumn 1985 1 of Treasury securities. This, in effect, implies that a very sizable percentage of the proceeds of sales of new Treasury securities to foreigners are being used to pay interest to existing foreign holders of Treasury debt. only 20 to 30 percent of net private savings flows despite the fact that in the cases of West Germany and the United Kingdom, the cyclical component of their deficits was larger than for the United States. If the current efforts at reducing Federal budget deficits are not successful, then even under fairly optimistic economic conditions, the annual cost of servicing the Federal debt by 1990 will be in the neighborhood of $210 billion...For every five dollars collected from the individual income tax, two dollars will go toward paying Federal net interest liabilities. The growth in Federal debt lies at the root of another dramatic development regarding the United States, and that, of course, is the sudden and sizable shift in the position of the United States from a net creditor to the rest of the world to a net debtor. The immediate cause of this development is, of course, the unprecedented current account deficits we are running, but as this audience would recognize, the underlying causes for those current account deficits are importantly related to the budget deficit via the interest rate, exchange rate nexus. Here too, orders of magnitude are so large that they can lose meaning, but the following provides some perspective: • Looking out over the next few years, if the current efforts at reducing Federal budget deficits are not successful, then even under fairly optimistic eco nomic conditions, the annual cost of servicing the Federal debt by 1990 will be in the neighborhood of $210 billion. That will mean that for every five dollars collected from the individual income tax, two dollars will go toward paying Federal net interest liabilities. Moreover, even if near-term deficits were reduced to levels consistent with the targets spec ified in the recent Budget Resolution, annual net interest payments by the Federal Government would still grow to $180 billion five years from now. • Federal debt relative to GNP, which had been on a pronounced downward trend over most of the post war period, is now rising very rapidly. Indeed, for 1985, Federal debt will amount to almost 40 percent of GNP—a rise .of more than twelve percentage points since 1981. • At this juncture, and short of worldwide economic conditions that would be most distasteful, it is dif ficult to foresee circumstances in which the foreign debt of the United States would not approach $500 billion by the end of the decade. Indeed, some would suggest that we would have to be quite lucky if that figure were not larger than $500 billion. In considering the possible implications of external debt of this size, there is at least a question as to whether foreigners will be eager to continue to accumulate dollar denominated assets of the amounts suggested at current, much less lower, rates of interest. For the private sector as a whole, the ratio of debt to GNP is at an unprecedented level and is still rising. Short of worldwide economic conditions that would be most distasteful, it is difficult to foresee circumstances in which the foreign debt of the United States would not approach $500 billion by the end of the decade. • Servicing $500 billion in external debt at roughly current interest rates could produce a $35 to $45 billion gap between our trade and current account deficits and would imply that even approaching current account balance will require not just a bal ancing of our trade account but moving the trade account into a sizable surplus position. • An even more alarming picture arises when we look at the deficit relative to our domestic savings flows. In 1984, for example, the deficit consumed twothirds of our net private domestic savings. While international comparisons are flawed, it is never theless noteworthy that in Japan, West Germany, and the United Kingdom, budget deficits consumed • The “Catch-22” of this situation, however, is that so long as our budget deficits are so large and our domestic savings so meager, we are vitally dependent on those same foreign savings flows which finance the current account deficit to finance our domestic activities including the budget deficit. At present, foreign savings flows are augmenting our net private domestic savings by a factor of 2 FRBNY Quarterly Review/Autumn 1985 more than one-third and are directly or indirectly financing half or more of the budget deficit. In summary, looking at the rate at which we are building debt in the Federal sector and looking at the closely related issue of the rate at which the United States is accumulating external debt, it is difficult to escape the conclusion that we are approaching or in uncharted waters. But, even that’s only part of the story since it does not take account of developments regarding debt accumulation in the private sector. Abstracting from internally generated equity, the 1984-85 period will, if current trends continue, see the net retirement of $150 billion of equity in the nonfinancial corporate sector—an amount which in nominal dollars exceeds the net issuance of equity by nonfinancial business over at least the entire post-Korean War period. To some extent, private sector debt accumulation has been overshadowed by events in the public sector. And, to some extent they have been muted by what, in my judgment, may be a false sense of security growing out of some statistics which, for example, suggest that consumer liquidity is relatively high and rising or that certain debt ratios for nonfinancial business have stopped rising, or are falling slightly. Takir\g those and other statistical indicators at face value, one could, perhaps, conclude that outside of the Federal Govern ment, all is reasonably well. Perhaps that is so, but I would suggest that a closer look at trends in the private sector may not justify that complacency. existing levels of debt in a less favorable economic and interest rate environment could prove very dif ficult. This is especially true since generalized financial indexation has shifted a sizable fraction of overall interest rate risk from the financial sector to the nonfinancial and household sectors. • Taking account of where we are in the business cycle, some measures of credit quality problems are disquietingly high. This is especially true, for example, for delinquency rates on home mortgages, and of the overall level of nonperforming loans in the banking system. • The recent growth in debt has been associated with a very rapid retirement of equity which, in turn, is importantly—but not exclusively— related to lever aged buyouts and the threats of hostile takeovers. For example, abstracting from internally generated equity, the 1984-85 period will, if current trends con tinue, see the net retirement of $150 billion of equity in the nonfinancial corporate sector—an amount which in nominal dollars exceeds the net issuance of equity by nonfinancial business over at least the entire postKorean War period. Given all that has happened regarding patterns of debt accumulation in recent years, it is not easy to capture the underlying reasons for these developments in a few paragraphs. In the case of the Federal sector, I believe that most would now agree that the problem is primarily one of a political nature. Thus, rather than rehashing the familiar elements of that situation, allow me to focus my commentary on the major factors which seem to lie behind developments in the private sector. I say that for several reasons including the following: • For the private sector as a whole, the ratio of debt to GNP is at an unprecedented level and is still rising. To be sure, the increase is not as pro nounced as for the Federal Government, but there is at least a question as to whether it is reasonable to assume there is that much more good quality debt relative to GNP today than there was a decade or two ago. • The recent spurt in private sector debt accumulation has, to a large extent, occurred on the upside of the business cycle and the downside of the nominal interest rate cycle and despite what are generally seen as relatively high real interest rates. Since it does not seem at all prudent to assume that the business cycle is a thing of the past, servicing even It would appear that at least some borrowers and their lenders are still assuming—consciously or subconsciously—that inflation will bail them out. To the extent that is true, it strikes me as a very bad bet. To some extent, recent developments regarding pri vate debt accumulation reflect longer-term trends. Among the longer-term factors, demographics are such that we now have a relatively heavy clustering of the population in age groups that are more prone to borrow. Similarly, a case can be made that a host of techno logical, institutional, and innovational factors ranging from credit cards to junk bonds are working in the direction of enhancing the accessibility to credit. So too, FRBNY Quarterly Review/Autumn 1985 3 a case can be made that the worldwide integration of money and capital markets broadens financing options and alternatives for many corporations at any given level of interest rates. These and other factors may be playing a role in the burgeoning rate of debt accumu lation but they don’t seem capable of fully explaining why the experience of the recent few years looks so different than earlier periods. At the margin at least, it would seem that still other factors must be at work. Let me suggest two or three factors that may help to further explain recent behavior. • It would appear that at least some borrowers and their lenders are still assuming—consciously or subconsciously—that inflation will bail them out. To the extent that is true, it strikes me as a very bad bet. For one thing, it makes an assumption about monetary policy that, from my perspective, is simply wrong. However, it’s a bad bet in a more funda mental way because renewed inflation would inev itably bring more instability, not less. Indeed, I don’t think it unreasonable to assume that even a sniff of a new outburst of inflation would produce a financial market response in interest rates that could be quite harmful to those with high debt service burdens. Reducing the budget deficit is central not only to establishing a better balance in the utilization of our domestic saving, but it is the only vehicle through which we can achieve an orderly reduction of our dependency on foreign savings while still leaving enough room to finance the domestic investment ultimately needed for economic growth. • It is possible that very intense competitive forces in the banking and financial sector are such that the pricing of loans and other debt obligations does not fully take account of differences in credit risk, thereby diminishing the rationing effects of the pricing mechanism for debt. • Financial innovation may be aiding and abetting the debt accumulation process in part by transferring the incidence of credit and interest rate risk in ways that may give rise to the illusion that such risks have been reduced or eliminated. • Innovational forces have also given rise to certain highly sophisticated financing techniques which are designed to take maximum advantage of certain 4 FRBNY Quarterly Review/Autumn 1985 features of the tax code—a tax code which has strong incentives for debt accumulation in the first instance. Highly leveraged buyouts are the obvious example, but sophisticated tax shelter devices— which by definition spur debt creation—are now readily available even to individual investors with relatively modest income levels. A casual reading of the book Funny Money which deals with the Penn Square debacle provides a number of amusing but tragic insights into how easily even sophisticated investors can get duped by sure fire "deals” of this nature. We must continue to resist the temptation that the solution to our debt accumulation problem lies with accepting a little more inflation. What’s interesting about those episodes in Funny Money is that they may be symptomatic of a cultural revolution about debt. Homeowners no longer burn the mortgage when it’s paid; they quickly get another, and preferably one which, in effect, requires no payment of principal; commercial real estate developers shun even minimal equity investments in new projects; corporate takeover specialists finance their activity by leveraging to the hilt; and in each of these cases somewhere there seems to be a financial institution that will eagerly oblige. In short, the factors that lie behind the rapid growth in debt in the U.S. economy represent a complex interaction of political, economic, technological, market, and attitudinal considerations that will not easily be reversed. Yet, common sense tells us that a continuation of recent trends is not sustainable over the long haul. Looked at in that light, the crucial question, of course, is how can we best go about the process of slowing the rate of debt accumulation in a way that maximizes the prospects for more balanced non-inflationary economic growth in the period ahead. From my perspective, the answer to that question lies in several closely related areas of public policy and private initiative, as follows: • First, and perhaps most essentially, we simply must do more to reduce the budget deficit in a timely and credible manner. The recently enacted budget res olution—if adhered to— is a positive step and pro vides a margin of breathing room in the near term. But, more needs to be done in a context in which the next steps may be even more difficult to achieve. Reducing the budget deficit is central not only to establishing a better balance in the utili zation of our domestic saving, but it is the only vehicle through which we can achieve an orderly reduction of our dependency on foreign savings while still leaving enough room to finance the domestic investment ultimately needed for economic growth. And, only with that need for foreign savings reduced can we bring about the orderly adjustment in our external deficits that is also so essential. • Second, we should continue to explore ways in which tax policy can be tilted in the direction of greater incentives for savings and equity invest ment. Indeed, the current tax codes—with acrossthe-board deductibility of interest and the de facto double taxation of profits—create powerful motives for debt accumulation by households and busi nesses alike. To the extent that situation can be altered somewhat in the direction of greater incen tives to save and to finance through equity, we will be that much better off. Indeed to the extent we can achieve that tilt in a context in which the deficit is also coming down in a decisive way, our pros pects for sustained growth will have been enhanced appreciably. Despite enormous competitive pressure that works in the opposite direction, managers and directors of individual financial institutions will have to more fully recognize that more conservative lending and funding policies are ultimately in their individual and collective interests. • Third, we must continue to resist the temptation that the solution to our debt accumulation problem lies with accepting a little more inflation. Indeed, and as I noted earlier, more inflation can only bring more instability and greater problems down the road. • Fourth, we must seek out ways to adapt the bank supervisory process to the realities of contemporary banking markets—markets in which many of the traditional sources of restraint have been eliminated by a combination of deregulation and technologicallydriven innovation. This effort must entail a general strengthening of the bank supervisory process but also the active exploration of approaches that can move in the direction of encouraging financial institutions to take on more liquid and less risky assets. The latter is one of the reasons why I am strongly attracted to the concept of seeking to take account of risk characteristics in the development and administration of capital adequacy standards for banking institutions. • Fifth, turning to the private sector, we must see a greater renewal of the precepts of prudence and discipline in the management of banking and financial institutions. Even now there is some evi dence to suggest that renewal is beginning to take hold as illustrated, for example, in the number of institutions that are maintaining capital positions well in excess of regulatory minimums. Yet, short term preoccupation with growth and quarterly earnings performance still seems unbalanced and misplaced. More generally, and despite enormous competitive pressure that works in the opposite direction, managers and directors of individual financial institutions will have to more fully recog nize that more conservative lending and funding policies are ultimately in their individual and col lective interests. In conclusion, the debt accumulation problem is a matter of concern. Some elements of it will be selfcorrecting but others will need an assist from public policy and from private initiatives. Those initiatives constitute something of an insurance policy—and a relatively inexpensive one at that—which can signifi cantly raise the probabilities that we can sustain an economic and financial environment conducive to growth without inflation. FRBNY Quarterly Review/Autumn 1985 5 Federal Tax Reform and the Regional Character of the Municipal Bond Market Of the various tax proposals that could affect the municipal bond market, reduction of marginal tax rates and repeal of state income tax deductibility require special attention. Analysts are aware that repeal of state income tax deductibility would increase the out-ofpocket, effective level of state taxation. They also know that lower federal marginal tax rates would reduce the value of federal tax exemption of municipal bonds. That analysis is incomplete, however, because of two important characteristics of the municipal bond market. First, most states impose taxes on the income their residents earn from bonds issued out-of-state. Any increase in effective state income taxes would raise the value of in-state bonds to investors and equivalently penalize borrowers who need funds from out-of-state. Second, because the majority of municipal bonds are bought by local investors, the effects of reducing the value of a bond’s federal tax exemption depend on how many investors are affected in each state. Current federal tax law fosters some uniformity in the municipal bond market by limiting the variations across states due to these two market characteristics. Repealing deductibility and establishing fewer brackets at lower marginal rates would remove these limits. They would raise interest costs for borrowers in some states and lower costs for those in other states. Though these are only two of many reform proposals that affect the The authors would like to thank Daniel Chall for his derivations of state tax formulas. 6 FRBNY Quarterly Review/Autumn 1985 municipal bond market, they are interesting because each state is affected differently.1 In attempting to identify how widely the effects of reform may vary across states, this analysis begins by describing how state tax laws contribute to the regional character of the bond market. The second section describes the role of demand for bonds by state resi dents relative to in-state borrowing needs. State tax laws and populations in each tax bracket are then analyzed to contrast the effects of current federal law with those of the most recent Administration proposals. The findings suggest that these proposals may have effects on the cost of borrowing that vary widely from one state to another. ’ Some other proposals may affect the bond market to a larger degree, but their effects should be roughly similar across all states. They would raise or lower interest rates about the same for one state as for another. But the overall combined effect of the other proposals is uncertain. Viewed in isolation, some may create upward pressures on yields across states while others may create downward pressures. For example, the proposed elimination of federal tax exemption on many types of revenue bonds may reduce supply and lower yields. At the same time, a reduced number of alternative tax shelters may increase the value of tax-exempt bonds, raise demand, and lower yields. However, eliminating special treatment of commercial bank investment in tax-exempt bonds is likely to move many banks out of the market, lower demand, and raise rates over time. On balance, it is difficult to know whether yields will rise or fall as a result. For detailed analysis of the influence of federal tax law on commercial bank investment in municipal bonds, see Allen J. Proctor and Kathleene K. Donahoo, “Commercial Bank Investment in Municipal Securities”, this Q uarterly R eview (Winter 1983-84). For approximations of possible effects on the average national level of interest rates, see Andrew Silver, "Three Aspects of the Administration’s Tax Proposal: Tax-Exempt Rates”, this Q uarterly R eview (Summer 1985). State tax laws and the favored treatment of in-state bonds The municipal bond market has a regional orientation for most borrowers. In general, local investors buy the bonds local borrowers issue and local market conditions determine their borrowing costs.2 There is also a more familiar national market, con sisting of a relatively small number of nationally rec ognized borrowers who regularly issue large volumes of bonds. Investors throughout the country buy and sell their bonds, and national market conditions determine their borrowing costs. One factor shared by municipal bonds in both markets is exemption from federal income taxes. Because no bond income needs to be set aside to pay federal taxes, investors are willing to accept lower yields than they would on investments subject to federal tax. The ratio of tax-exempt to taxable yields is often used to identify the federal tax bracket of the marginal investor in the national market. Outside the national market, state taxation of munic ipal bonds becomes an important reason for the cost of borrowing to vary from one state to another. Puerto Rican municipal bonds are not taxable in any state, but 38 states presently impose some form of tax on other municipal bonds. Of the remaining 12, seven have no tax on any form of income and five impose no taxes on municipal bond income (Table 1). Thirty-five of the states that tax municipal bond income use their tax laws to create special preferences for in-state borrowers. In-state bonds are tax-exempt while out-of-state bonds are not. For example, an investor who lives in a state with tax preferences earns $900 in annual aftertax income from a $10,000 in-state bond paying a 9 percent yield. If the state income tax is 5 percent, an equivalent out-of-state bond would provide only $855 of income after $45 in state taxes was paid. To return the same aftertax income as the in state bond, the outside borrower must offer a resident investor a before-tax yield of 9.47 percent.3 This pref erence creates an incentive for borrowers to sell their bonds in th e ir home states. The preference also encourages residents to switch from out-of-state bonds to in-state bonds of equivalent value. The primary reason for creating tax barriers against outside borrowers is to improve the balance of supply 2For a discussion of the regional and national segments of the municipal bond market, see Robert Lamb and Stephen P. Rappaport, Municipal Bonds: The Comprehensive Review of Tax-Exempt Securities and Public Finance (1980), pages 27-50. ’ Local income taxes are not considered in this study. These taxes will enlarge the basis point disadvantage placed on out-of-state bonds. Factors other than yield will also affect an investor’s decision to buy out-of-state bonds: diversification, familiarity with the borrower, credit risk, etc. and demand between resident borrowers and investors. By making out-of-state entry into their markets more expensive, states hope to increase the demand for in state bonds among resident investors. If demand for municipals by residents is large enough to meet bor rowing needs, then in-state borrowers may be able to sell their bonds exclusively to residents and achieve the maximum reduction of borrowing costs that the tax barriers permit. If demand by resident investors remains too small to absorb the supply of in-state bonds, despite the state’s encouragement of in-state investment, bor rowers will need to attract investors from outside the state. A municipal borrower who goes out of state to find enough funds, however, must compete in other bor rowers’ home markets and overcome whatever tax Table 1 Effective State Income Taxes on Municipal Bonds No state tax* Type of security Out-of-state municipal .. In-state municipal .. Number of states ... All municipals exem ptf No tax preference for in-state bonds No municipals exempt^ Tax preference for in-state bonds Only in-state municipals exempt§ — — S (1 - F) S (1 -F ) — — S (1 —F) — 7 5 3 35 Key: F= Federal marginal income tax rate. S= State marginal income tax rate. — = No income tax. The exact tax preference for in-state bonds depends on state tax rules (Appendix 1) and is generally equal to the state tax rate reduced by the federal deduction of state taxes. ’ Alaska, Florida, Nevada, So. Dakota, Texas, Washington, and Wyoming. flndiana, Nebraska, New Mexico, Utah, and Vermont. ^Illinois, Iowa, and Wisconsin. Iowa exempts only Iowa State Board of Regents bonds and Wisconsin exempts only Housing Authority bonds. §Alabama, Arizona, Arkansas, California, Colorado, Connecticut, Delaware, Georgia, Hawaii, Idaho, Kansas, Kentucky, Louisiana, Maine, Maryland, Massachusetts, Michigan, Minnesota, Mississippi, Missouri, Montana, New Hampshire, New Jersey, New York, No. Carolina, No. Dakota, Ohio, Oklahoma, Oregon, Pennsylvania, Rhode Island, So. Carolina, Tennessee, Virginia, and West Virginia. Colorado, Kansas, Ohio, and Oklahoma tax some types of in-state bonds. Source: Hueglin and Ward, op. cit. FRBNY Quarterly Review/Autumn 1985 7 barriers may be imposed. The borrower needs an underwriter who has a broad and strong broker network that can convince individual investors to buy unfamiliar, out-of-state bonds. The bonds must also offer a taxable yield that provides at least the same aftertax return the investor can earn from untaxed in-state bonds. For the 15 states without tax preferences, the advantage of borrowing from resident investors and the importance of resident demand and supply is less clear (Table 1). Resident investors in these “free access” states receive the same tax treatment on in-state and out-of-state bonds. Outside borrowers, therefore, face no barriers to seeking resident investors, and in-state borrowers must always compete against borrowers from the other 14 free-access states. This generally will raise the in-state cost of borrowing. Moreover, changes in resident demand for in-state bonds may not be sym metrical when there are nationwide changes in demand for municipal bonds. Out-of-state borrowers have access to resident investors to try to shift part of any increase in demand away from in-state borrowers. At the same time, they may shift any reduction of demand onto in state borrowers by intensified bidding for resident investors. Current effectiveness of tax preferences Even though use of tax preferences is widespread, their current importance depends on the size of the tax bar riers and the need for borrowers to cross the barriers. Federal tax law plays an important role in each. Federal deductibility of state income taxes lowers in state bond demand by reducing the out-of-pocket cost of state taxes. This reduction occurs because each dollar of state income tax is partially offset by a reduc tion of federal taxes for taxpayers who deduct state income taxes. Instead of a combined federal (F) and state (S) tax rate of F + S, taxpayers face a rate of F - FS + S, where FS represents the federal tax reduction from deduction of state taxes. The effective out-of-pocket cost of state taxes is S - FS, which is restated as S(1 - F ) in Table 1. For example, an investor in the 25 percent federal tax bracket, who faces a 5 percent state tax on a $10,000 out-of-state bond yielding 9 percent, can use deduct ibility to reduce his federal taxes by one-fourth of his $45 state tax bill. Thus, he pays a $33.75 state tax on the income from his out-of-state bond. Deductibility increasingly blunts the effectiveness of tax barriers as the federal tax bracket increases. At the top federal bracket of 50 percent, for example, state taxes are reduced by half. If the resident investor with the $10,000 bond and $45 state tax bill were in this bracket, his effective state tax would be only $22.50. This federal offset also limits the yield an out-of-state 8 FRBNY Quarterly Review/Autumn 1985 borrower would have to offer a resident to equal the aftertax return of a comparable in-state bond. In the example above, the out-of-state borrower would have to pay a top-bracket investor 9.23 percent to equal the aftertax return on a 9 percent in-state bond. However, this increase of 23 basis points is lower than the 47 basis points the outside borrower would have to pay without federal deductibility. Estimates of the size of tax preferences in each state show that current law with federal deductibility results in relatively modest barriers to outside borrowers.4 Using comprehensive measures of effective state tax rates, Steven Hueglin and Karyn Ward calculate the aftertax return of equivalent bonds in each of the states with tax preferences. In about one-third of the states, the aftertax return of an equivalent outside bond is less than 30 basis points below an in-state bond. In all but five states, state taxation lowers the return on out-of-state bonds by less than 50 basis points. The states with larger barriers to outside borrowers are Delaware (61), Minnesota (89), Montana (52), New York (63), and West Virginia (71).s Whether borrowers need to cross these tax barriers depends on the demand for their bonds in their home states. Each state’s tax schedule and specific tax rules provide a unique schedule of effective tax rates by income bracket. Based on these tables and the actual interest rates on municipal, Treasury, and corporate bonds, it is possible to specify those investors who would prefer in-state municipal bonds to all other bonds. This pool of potential investors can be characterized as all taxpayers above a certain income tax bracket, which varies by state. In California in 1984, for example, in-state municipal bonds provided the highest average aftertax returns for residents with taxable incomes above $24,600. Based on the tax formulas in Appendix 1, the average Treasury bond yielding 12.46 percent and the average medium grade corporate bond yielding 14.14 percent gave a California investor in that tax bracket aftertax returns of 9.35 percent and 10.07 percent, respectively. By com parison, the average California bond in 1984 yielded 10.11 percent. At higher tax brackets, the superiority of in-state municipals would widen. At lower tax brackets, Presum ably all municipal bond investors lower their effective tax rates through deductibility. Seventy percent of all married taxpayers filing joint returns with taxable incomes over $30,000 deduct state and local income taxes. This income level coincides closely with the minimum taxable income for resident investors in most states. *Steven Hueglin and Karyn Ward, G u id e to S ta te a n d L o c a l Taxation o f M u n ic ip a l B onds (1981). Their calculations are based on a 9 percent coupon bond selling at par using approximations of the formulas presented in Appendix 1. They also include personal property taxes for states that have such taxes. Since they performed their calculations, Connecticut has introduced taxation on out-ofstate bonds. corporate bonds would have a higher aftertax yield than both California municipals and Treasury bonds. In other states with different tax schedules, rules, and average yields, the aftertax return on in-state bonds becomes superior at different income levels. These brackets are presented in Table 2 (column 1) based on 1984 tax laws and interest rates. In Alabama, for example, the average in-state yield becomes superior to other yields above the $35,200 income bracket. The need for borrowers to go outside the state to find sufficient investors can be approximated by the ratio of total municipal borrowing in the state and the number of potential resident investors.6 A high dollar value per investor suggests a high probability that borrowers in that state often cross state lines and possibly encounter tax barriers. This may occur because the state has few high-income residents to demand the bonds or because its borrowing needs create a relatively large supply of bonds. Conversely, a low value suggests that a state is able to function as a self-sufficient market in which all supply is taken up by resident demand. This may occur because demand is high owing to a large high-income population or because supply is low owing to relatively limited borrowing needs. The estimates of bonds issued per potential resident investor range from a high of $60,700 in Wyoming to a low of $2,800 in Ohio and Indiana (Table 2, column 3). There is no particular level of per capita borrowing at which a state becomes self-sufficient. However, results from a study by Kidwell, Koch, and Stock suggest that at this time the majority of states are self-sufficient.7 The •The number of investors is approximated by the number of federal tax returns above the minimum taxable income level for each state. For this article, the alternative investments available to an investor are limited to U.S. Treasury bonds and corporate bonds. For other types of investments it is assumed that other factors, such as capital gains taxes or depreciation rules, are more important in calculating return than are income taxes, which are the focus of this article. See Appendix 1 for a discussion of how aftertax returns are calculated for each type of bond. An alternative measure of the ability to sell exclusively to residents is the ratio of dollars issued to the aggregate income of potential resident investors. Use of this measure does not alter the results appreciably. 7David Kidwell, Timothy Koch, and Duane Stock, "The Impact of State Income Taxes on Municipal Borrowing Costs", N a tio n a l Tax Jo u rn a l 37 (Decem ber 1984) pages 551-562. Their study examines yields on general obligation bonds of less than $5 million which were bid competitively in 1980. The study finds that tax preferences on average are successful in reducing the cost of borrowing for in state borrowers relative to outside borrowers. Significantly, however, the average reduction is a fraction of the value of the tax preferences. This partial effect may occur if the marginal investors for some of the bonds are not state residents and therefore do not benefit from tax preferences. In that sense, these results confirm that, while some municipal bonds are sold in-state (where tax preferences lo w er the cost of borrowing), a significant proportion of municipal bonds are sold out-of-state, where tax preferences raise the cost of borrowing. estimates in Table 2, then, are one way to sort out which states lower their costs through tax preferences by being self-sufficient and which see their costs raised because they must cross other states’ tax barriers. About 30 of the 35 states with tax preferences may have enough resident investors to be self-sufficient for in-state borrowing needs if around $10,000 of borrowing per investor were the cutoff point. These^ may be the states, then, that are able to lower their borrowing costs by imposing taxes on out-of-state bonds. On the other hand, Michigan, New Jersey, North Carolina, and North Dakota may not benefit from their taxation of out-of-state bonds. Borrowers in these four states issue much more than $10,000 per resident investor. They are more likely, therefore, to require additional investors from outside the state. Most of the 15 states which do not protect their in state borrowers have low borrowing needs relative to their investor pool. Their borrowers are probably able to avoid the increased costs of crossing the tax barriers of other states. In sum, under present law, demand and supply con ditions in most states do not indicate that a great deal of interstate borrowing is occurring in the municipal bond market. Local borrowing from local investors appears sufficient to satisfy financing needs in most states. For the relatively few borrowers who may depend on outof-state sales, the effective state taxes they may encounter seem to be relatively modest. Tax reform and its effect on interstate competition for investors Federal tax reform has important effects on interstate differences in the municipal bond market. Resident demand for in-state bonds is sensitive to any change in federal tax rates, and the size of tax preferences is sensitive to any change in federal deductibility of state income taxes. Most proposals for federal tax reform will change at least one of these provisions. The remainder of this article uses the President’s Tax Proposals to the Congress for Fairness, Growth, and Simplicity (Treasury II) to illustrate what the effects of these two provisions would be on regional municipal bond markets and why the effects would vary widely across states. Increased need to borrow out-of-state The federal tax reform proposal is structured so that tax rate cuts are not the same for every state pool of potential resident investors. Treasury II proposes marginal tax rates of 15 percent for incomes to $29,000, 25 percent for incomes to $70,000, and 35 percent for incomes over $70,000. In states like Colorado, present marginal investors in the resident pool have taxable incomes under $30,000. For them, FRBNY Quarterly Review/Autumn 1985 9 the marginal tax rate will remain unchanged at 25 percent. In other states like Alabama, the marginal investor at current interest rates has taxable income of $35,200. The proposal reduces that investor’s tax rate from the current level of 33 percent to 25 per cent. And in states like New Jersey where the tax able income of the marginal investor is $45,800, the marginal tax rate declines from 38 to 25 percent. These lower tax rates will reduce the appeal of municipal bonds relative to taxable bonds. Many of today’s marginal investors will drop out of the market, causing demand for in-state bonds to decline and the minimum income level of the rem aining potential investors to be higher. Estimates of these new income levels are presented in Table 2 (column 2) for current rates of interest. For most states, the return on in-state municipal bonds will no longer appeal to residents earning less than $70,000. The current before-tax yield spread between in-state municipals and taxable bonds is too wide for most residents in the proposed middle tax bracket. In only nine states (Arkansas, California, Col orado, Delaware, Hawaii, Idaho, Maryland, Montana, and Oregon) do state and federal taxes on Treasury and corporate bonds combine to make current in-state municipal yields attractive to the middle-bracket investor earning between $29,000 and $70,000. Estimates of the percentage of current potential investors who will continue to demand in-state munic ipals are presented in Table 2 (column 5). The nine states where m iddle-bracket investors are likely to remain in the market at current yields should face only Table 2 State Characteristics of the Regional Municipal Bond Market State A la b a m a .................................... ............. Alaska ........................................ ............. Arizona ...................................... ............ Arkansas ................................... ............. California ................................... ............. Colorado ................................... ............ Connecticut .............................. ............. Delaware ................................... ............. Florida ........................................ ............. G e o rg ia ...................................... ............ Hawaii ........................................ ............. Idaho ......................................... ............. Illin o is ......................................... ............ Indiana ...................................... ............ Io w a............................................ ............ Kansas ...................................... ............. Kentucky ................................... ............. Louisiana ................................... ............ Maine ......................................... ............. Maryland ................................... ............. Massachusetts ......................... ............ Michigan ................................... ............ Minnesota ................................. ............ Mississippi ................................ ............. Missouri .................................... ............. Montana .................................... ............. Nebraska ................................... ............. Nevada ...................................... ............. New Hampshire ........................ ............. New Jersey .............................. ............ New Mexico .............................. ............ New York ................................... ............ No. C a rolin a .............................. ............ No. Dakota ................................ ............. O h io ............................................ 10 Minimum tax bracket of resident investors* In dollars 1984 law Proposed law (D (2) 35,200 29,900 29,900 29,900 24,600 24,600 50,000 24,600 32,500 29,900 24,600 29,900 45,800 35,200 45,800 35,200 35,200 29,900 35,200 29,900 35,200 45,800 35,200 29,900 29,900 24,600 29,900 35,200 35,200 45,800 35,200 29,900 45,800 35,200 35,200 FRBNY Quarterly Review/Autumn 1985 70,000 70,000 70,000 29,000 29,000 29,000 70,000 29,000 70,000 70,000 29,000 29,000 70,000 70,000 70,0001 70,000 70,000 70,000 70,000 29,000 70,000 70,000 70,000 70,000 70,000 29,000 70,000 70,000 70,000 70,000 70,000 70,000 70,000 70,000 70,000 Dollar borrowing per potential resident investor^ In thousands of dollars 1984 law Proposed law (3) (4) 8.7 24.8 8.1 3.3 7.0 8.7 11.0 7.8 11.1 9.5 6.1 3.9 15.8 2.8 19.2 7.0 8.8 8.4 4.2 3.7 4.6 13.7 7.5 5.8 5.3 10.5 5.3 7.1 3.2 18.4 8.9 5.3 21.7 16.9 2.8 79.1 116.4 66.9 3.3 7.0 8.7 18.4 7.8 70.2 77.7 6.1 3.9 27.2 28.4 33.51 53.3 78.7 61.1 39.8 3.7 33.5 24.4 63.8 49.3 44.3 10.5 44.6 52.2 27.7 32.1 73.7 34.2 37.1 136.1 25.6 Retention of potential resident investors under proposed law} In percent (5) 11.0 21.3 12.1 § § § 59.6 § 15.8 12.2 § § 58.2 9.9 57.11 13.1 11.2 13.7 10.6 § 13.8 56.0 11.8 11.7 11.9 § 11.8 13.6 11.5 57.2 12.0 15.6 58.5 12.4 11.1 a small change in demand. All other states may face a significant loss of investors.8 The effect of these changes on the cost of bor rowing in each state depends on how much demand falls short of local borrowing needs. Table 2 (column 4) presents estimates of the amount of borrowing per investor if 1984 borrowing needs continue. Virtually all the 41 states losing middle-bracket investors will have per capita borrowing levels that exceed current levels. New York provides an illustration of the consequences of losing a large number of investors in the critical •The current spread between municipals and taxable bonds is larger at short maturities than at longer-term maturities. The loss of in-state demand will be largest at the maturities with the largest spreads along the future yield curve. $29,000 to $70,000 range. New York borrowers currently issue about $5,000 in bonds per potential resident investor annually. This is low, but m iddle-bracket investors represent all but 15 percent of the investor pool. This is the very group that is likely to drop out of the market at current yields. If New York borrowers were to lose middle-bracket investors, their sales to resident investors would need to average $34,000 per potential investor. At present, only two states issue such a large amount of debt per capita. The reduced pool of investors may not absorb so much debt at current yields. Evidence cited earlier suggests that the states with per capita borrowing above $10,000 may currently rely on out-of-state investors for at least part of their borrowing needs. Short of reducing their future bond issuance substantially, borrowers in the Table 2 State Characteristics of the Regional Municipal Bond Market, continued State Oklahoma ............................ ................. Oregon ................................. ................. Pennsylvania ......................... ................. Rhode Island ........................ ................. So. Carolina ............................................ So. Dakota ............................ ................. Tennessee ............................................... Texas...................................... ................. U ta h .......................................................... Vermont.................................. ................. Virginia .................................. ................. W ashington............................ ................. W. Virginia ............................................... Wisconsin ............................................... Wyoming ............................... ................. Minimum tax bracket of resident investors* In dollars 1984 law Proposed law (1) (2) 35,200 29,900 35,200 35,200 35,200 35,200 35,200 35,200 35,200 45,800 35,200 29,900 35,200 35,200 45,800 70,000 29,000 70,000 70,000 70,000 70,000 70,000 70,000 70,000 70,000 70,000 70,000 70,000 70,000 70,000 Dollar borrowing per potential resident investorf In thousands of dollars 1984 law Proposed law (3) (4) 5.6 4.7 5.3 11.2 10.1 11.5 5.5 8.3 21.2 30.6 4.1 3.5 4.4 3.6 60.7 Retention of potential resident investors under proposed lawt In percent 38.5 4.7 45.5 95.7 94.5 115.8 46.7 53.2 215.0 55.0 28.4 28.3 46.0 36.3 104.4 (5) 14.4 § 11.6 11.7 10.7 9.9 11.8 15.6 9.9 55.6 14.3 12.3 9.7 10.0 58.1 *The minimum taxable income in 1984 at which the Public Securities Association estimates of the average net interest cost on in-state municipal bonds exceeds both the aftertax return on ten- and 20-year Treasury bonds (whose 1984 yields averaged 12.46 percent) and the aftertax return on Baa corporate bonds (whose 1984 yields averaged 14.14 percent). See Appendix 1 for the formulas used to calculate combined federal and state income taxes. State and federal tax schedules are available from the authors on request. Use of narrower yield spreads in the calculations would result in lower minimum income levels Calculations under the proposed law take into account both revised income tax brackets and repeal of federal deductibility, except for Iowa (see footnote below). fFor states with minimum taxable income levels up to $35,200 the number of potential investors is approximated by the number of federal returns with adjusted gross income (AGI) above $30,000. For states with minimum taxable income levels of $45,800 or $50,000 the proxy is the number of returns with AGI above $50,000. For a taxable income level of $70,000, the number of returns with AGI over $70,000 is computed as all returns above $100,000 AGI and one half the returns between $50,000 and $100,000 AGI. These estimates assume that 1984 levels of borrowing continue. Some other tax proposals may reduce future borrowing from current levels. tThe estimated number of potential resident investors under the proposed law as a percentage of current potential resident investors. §Virtually all potential resident investors will be retained. I lf deductibility is repealed, the spreads used in the calculations are too large for in-state municipals to be attractive to residents at any income level. Therefore, the effect of repeal of deductibility is not reflected here. Sources: Public Securities Association; Hueglin and Ward, op. c i t Internal Revenue Service, Statistics of Income; Advisory Commission on Intergovernmental Relations; and Federal Reserve Bank of New York staff estimates. FRBNY Quarterly Review/Autumn 1985 11 majority of states, therefore, would have two options.9 • They could increase yields by enough to induce the remaining resident investors to increase their holdings of in-state bonds. • They could sell their bonds out-of-state and pay premium yields to overcome the tax barriers other states may impose. Table 3 (column 1) presents estimates for selected states of the increased yields necessary to replace the lost investors. For states losing investors, the estimated increases range from 4 to almost 60 basis points. For example, for New York borrowers to sell all their bonds to the remaining resident investors, they would need to increase the average yield by an estimated 46 basis points over the 1984 average interest cost of 9.04 per cent reported by the Public Securities Association. In dollar terms, this increased yield would raise the debt service on a $10 million, 20-year bond issue by $920,000 over the life of the issue. An important reason some states may need larger increases in yields than others is the difference in the share of resident demand for in-state bonds which middle-bracket residents now represent. Appendix 2 presents a method for estimating these shares. In states with the largest estimated cost increases, middle-bracket residents currently represent a dispro portionately large share of demand compared with tppbracket residents. To replace middle-bracket demand, the remaining top-bracket investors must be induced by large increases in yields to raise the share of their income being invested in local bonds. By contrast, states in which top-bracket residents already account for most resident demand would have an easier time replacing their middle-bracket resident investor pool. For example, even though middle-bracket residents comprise about 90 percent of Utah’s pool of potential resident investors, they have only an estimated 73 percent of the income of the pool. Utah may have to give only a 4-basis-point increase in yields to convince its top-bracket resi dents to invest enough additional income in local bonds. As an alternative, borrowers may try to attract outof-state investors. In outside markets they will have to compete with more borrowers, some of whom are facing the same problem. In addition, they may need to attract investors from states that tax the income on out-of-state bonds. Repeal of federal deducti bility of state income taxes will have important •A reduction of borrowing may occur in some states as a result of proposed restrictions on certain types of municipal bonds. Data are not available to permit estimation of possible reductions by state. 12 FRBNY Quarterly Review/Autumn 1985 effects on their cost of going out-of-state. Increased barriers against out-of-state borrowing Repeal of federal deductibility of state income taxes would remove the moderating role of federal tax law on state tax preferences. Effective state taxes on out-of-state bonds would rise, placing outside bor rowers at a much greater yield disadvantage than they currently face relative to in-state borrowers. Estimates of the increased size of these preferences are shown in Table 3 (column 2) for selected states. Since states differ in their tax rates and rules, repeal of federal deductibility would have different effects across states on the value of tax preferences. For example, for a New York resident, repeal of deductibility would reduce the aftertax return of an outof-state municipal bond by 35 basis points.10 An outside borrower would have to increase the before-tax yield it pays by at least that much before it could compete with comparable New York borrowers for New York investors. This increase comes in addition to the 63-basis-point disadvantage out-of-state borrowers currently face in attracting New York residents.11 Some in-state borrowers in the 35 states with tax preferences may benefit from the increased barriers against outside borrowers. The increased value of state tax exemption may allow some in-state borrowers to reduce the yields they offer to residents. Residents who now hold out-of-state bonds may also replace some of them with in-state bonds and soften the effect of the loss of middle-bracket investors. Combined effects of federal changes The majority of municipal bonds are already sold on a regional basis in the United States. Revision of federal tax rates and repeal of deductibility would reinforce and possibly strengthen this local orientation of municipal financing. Repeal of deductibility would increase the incentive for borrowers to rely exclusively on resident demand for their bonds. At the same time, the possible loss of middle-bracket demand because of reduced federal tax rates would create a need for more intensive regional marketing of bonds in order to ensure enough resident investors for current borrowing needs. Self-sufficiency in financing local borrowing with local investment, however, will be far easier for some states than for others. The combined effects of fed eral tax reduction and repeal of deductibility divide the states into three classes according to the 10New York City residents will be affected to a greater extent because they also pay local income taxes on out-of-state bonds. "Hueglin and Ward, op.cit. Table 3 Possible Effects of Personal Income Tax Reform on In-State Borrowing Costs In basis points State Increased tax Increased cost of barriers against in-state borrowing* Qut-of-state borrowerst Alabama ................ Arkansas .............. California .............. Delaware .............. Florida ................... Hawaii ................... Indiana ................. Kentucky .............. Maryland .............. New York .............. O h io ........................ Oregon ................. Texas...................... U ta h ........................ Wisconsin ............. 14 0 0 0 17 0 59 19 0 46 54 0 25 4 39 15 19 13 19 0 21 0 18 20 35 17 27 0 0 0$ ‘ The increase in in-state borrowing costs necessary to maintain current resident demand if federal tax rates become 15 percent for incomes to $29,000, 25 percent for incomes to $70,000, and 35 percent for incomes over $70,000. fThe decrease in a resident investor’s aftertax return on an out-of-state bond relative to an equivalent in-state bond if federal deductibility of state income taxes is repealed. }:The repeal of federal deductibility will reduce the resident investor’s aftertax return on both in-state and out-of-state bonds by about 30 basis points. Because this state taxes both in-state and out-of-state municipal bonds, however, the repeal of deductibility will not affect the spread between the two types of bonds for a resident investor, A similar effect will occur in Iowa and Illinois. Source: Federal Reserve Bank of New York staff estimates. probable future cost of financing public projects: • states which are most likely to face increased borrowing costs because of a large decline in middle-bracket demand and an absence of tax barriers to discourage residents from financing outof-state projects; • states that are most likely to become more auton omous with reduced borrowing costs because of a continued large potential resident investor pool and increased tax barriers to discourage out-of-state investment; and • states that may become more autonomous but with varying changes in borrowing costs because a reduced resident investor pool will face increased barriers to investing out-of-state. The 15 states without tax preferences will be the markets of choice for borrowers from out-of-state who need to replace their lost middle-bracket investors. The increased number of borrowers competing for a reduced investor pool may create substantial pressures on bor rowers to raise yields. For example, Texas borrowers may need to increase yields by an estimated 25 basis points in order to induce top-bracket resident investors to replace the demand of middle-bracket residents. If more out-of-state borrowers also try to attract investors in this state, the larger supply may force yields even higher for in-state bor rowers. This effect could be limited if tax preferences were introduced.12 By contrast, nine states would encounter no loss of resident demand and their protection from outside competition would increase. For example, Oregon bor rowers would increase their yield advantage over out side competition by an estimated 27 basis points while their borrowing needs would remain at the low level of $4,700 per resident investor. One consequence is that they might be able to reduce the yields they offer residents. Twenty-six states may encounter the third class of effects: reform would increase the benefits of financial self-sufficiency at the same time that it would erode their ability to be self-sufficient. New York best represents this conflicting situation. In-state borrowers would be pro tected from outside competition for funds by one of the largest increases in tax preferences for in-state resident investment. At the same time, the predominance of middle-bracket residents in the New York investor pool would cause one of the largest decreases in resident demand. If the latter effect is larger, as estimated in Table 3, enhanced tax barriers would be of little benefit, and local borrowers might need to go out-of-state. They would have to find new markets, introduce unfamiliar New York local bonds to new investors, and possibly pay high enough yields to offset out-of-state taxation. A final issue in evaluating federal tax reduction and repeal of deductibility is the effect of increased reliance on regional municipal bond markets. Under current law, states with large borrowing needs but relatively small high-iqcome populations can seek investors in other states usually at little additional cost. These tax pro- 12The benefits of introducing tax preferences would be especially large in Wisconsin, Iowa, and Illinois which may lose resident demand as a result of each federal tax proposal. These states currently have no tax preferences because in-state bonds are taxed at the same rate as out-of-state bonds. Uniquely for them, repeal of deductibility would reduce resident aftertax returns on in-state bonds—by as much as 30 basis points in Wisconsin. Exemption of in-state bonds would prevent this effect and limit the problem to the replacement of middle-bracket demand. FRBNY Quarterly Review/Autumn 1985 13 posals would encourage states to tax out-of-state investment and to solve their financing needs more completely in local markets. Because of the variety of state tax laws and the diverse abilities of states to be financially self-sufficient, however, not all regional mar kets would fare equally well. Allen J. Proctor and Julie N. Rappaport Appendix 1: State Tax Formulas This appendix presents the formulas used to calculate effective state and federal income tax rates on municipal, corporate, and Treasury bonds. These formulas are applied to taxable bond yields to determine the minimum income tax bracket for potential resident investors in each state (Table 2). They are also used to calculate the effect of repeal of deductibility on aftertax returns of outof-state municipal bonds (Table 3, column 2). Tax rates on fixed income securities for states can be divided into six groups on the basis of their deductibility formulas. The formulas use the following symbols: F = Federal marginal income tax rate S = State marginal income tax rate d = Deductibility of state and local income tax from the federal tax base: d = 1 under 1984 tax law d = 0 under proposed federal tax law C = Effective combined federal and state income tax rate on corporate bonds T = E ffective com bined income tax rate on Treasury bonds M = Effective combined income tax rate on outof-state municipal bonds Under current law, taxpayers who itemize on their federal returns may deduct their state and local income tax from their federal taxable income, for states that impose a state tax. In those states that do not, only the federal tax rate, F, applies to both Treasury and corporate bonds, and the effective tax rate on all municipal bonds is zero. These states are Alaska, Florida, Nevada, South Dakota, Texas, Washington, and Wyoming. For many states, deduction of state and local income tax from federal taxable income reduces the effective state tax rate. These are their formulas: C = F + [S(1 —dF)] T = F M = S(1 - dF) These tax formulas apply to Arkansas, California, Con necticut, Delaware, Georgia, Idaho, Illinois, Indiana, Maine, Maryland, Massachusetts, Michigan, Mississippi, New Hampshire, New Jersey, New Mexico, New York, North Carolina, Ohio, Oregon, Pennsylvania, South 14 FRBNY Quarterly Review/Autumn 1985 Carolina, Tennessee, Virginia, West Virginia, and Wis consin. For Illinois and Wisconsin, the formula is the same for out-of-state municipals and in-state municipals. Some states seek to lessen the tax burden further by also allowing the deduction of federal income taxes from state taxable income. For Alabama, Arizona, Iowa, Kentucky, Louisiana, Montana, and Oklahoma these tax formulas apply: C = F + [[(1 —dF)(S - FS)]/(1 -d F S )] T = F - [[(1 -dF)(FS)]/(1 -d F S )] M = [S(1 -dF )]/(1 - dFS) For Iowa the formula for out-of-state municipals also applies to in-state municipals. In other states, however, the additional tax savings from state deductibility of federal taxes are reduced because all state and local income taxes that were subtracted from the federal tax base must be added back into the state tax base. As a consequence, Colorado, Kansas, Minnesota, Missouri, North Dakota, and Utah use these formulas: C = F + [[(1 - d F )( S - F S )]/(1 - d F S -d S )] T = F - [[(1 -dF)(FS)]/(1 -d F S - d S ) ] M = [S(1 ~dF)]/(1 - d F S -d S ) In some states, income tax is calculated as a per centage of federal income tax. For Nebraska, Rhode Island, and Vermont, one formula applies to both cor porate and Treasury bonds: C = T = [F(1 + S)]/(1 +dFS) Tax treatment of municipal bonds differs among the three. Since Rhode Island exempts only in-state munic ipals from income tax, it has a separate tax formula for out-of-state municipals: M = [FS(1 - dF)]/(1 +dFS) On the other hand, Nebraska and Vermont exempt all municipal bonds, so that the effective combined tax rate on these securities is zero. Finally, in Hawaii, state income tax is deductible from the state income tax base as well as from the federal tax base. As a result, Hawaii has unique tax formulas: C = F - [[dFS/(1 + S)] + [S /(1+S )]] T = F M « [S/(1 +S)] - [dFS/(1 +S)) Appendix 2: Estimating Resident Demand This appendix summarizes the methodology for esti mating the demand fo r in -sta te bonds by resident investors. It also explains the calculation of the interest rate effects presented in Table 3 (column 1). In order to estimate the aggregate demand of potential investors in a given state two problems must be overcome. First, data on aggregate state income by bracket are provided for adjusted gross income (AGI). In contrast, taxable income is the basis for determining the minimum income of a potential investor. Consequently, the minimum tax able income levels in Table 2 must be converted to AGI. The initial AGI estimate is based on the ratio of AGI and ta xa b le incom e fo r each state and the ratio nationally for each AGI bracket. This estimate is further adjusted by the average amount of state and local income tax deducted by the average taxpayer at that level of AGI. The second problem occurs in estimating the aggre gate AGI of residents above this minimum level. Internal Revenue Service (IRS) data on state aggregate AGI by income level use bracket ranges that are larger than the range of most of the income levels examined in this study. As a result, interpolating aggregate income within the published income brackets requires estimating an income distribution function for each state using the following procedure. Based on IRS data on the number of returns and the value of income in each AGI bracket, we plotted two cumulative logarithmic distribution functions for each state: the cum ulative percentage of returns by AGI bracket and a Lorenz curve of cumulative percentage AGI and cumulative percentage returns. We located the estimated minimum AGI levels along each distribution function with a cubic spline function and then converted the results into the total state AGI above each minimum AGI level. The aggregate AGI of resident investors above the minimum taxable income level is approximated under 1984 law and the proposed law. The change in aggre gate income due to the proposals is adjusted for the assumption that 70 percent of the residents deducted state income tax from federal taxable income and that they invested an average of one percent of their gross income in municipal bonds each year. This income reduction is divided by bond issuance in each state to approximate the percent change in demand for in-state bonds. Using an interest elasticity of 1.27, the percent change in net interest cost is calculated. The value in basis points is based on the 1984 average net interest cost for each state estimated by the Public Securities Association. The elasticity estimate is taken from Patric Hendershott and Timothy Koch, “An Empirical Analysis of the Market for Tax-exempt Securities”, Monograph Series in Finance and Economics, New York University, Monograph 1977-4. For a discussion of using cubic spline interpolations of income distributions, see Christine Cumming and Roger Kubarych, "The Economic Effects of the Tax Deductibility of Interest”, Nominal and Real Interest Rates: Determinants and Influences, Bank for International Settlements (1985). FRBNY Quarterly Review/Autumn 1985 15 Recent Instability in M1’s Velocity The behavior of M1’s velocity during the 1980s has been remarkably different from the 1970s. After increasing about 3.5 percent per year during the 1970s, M1’s velocity has shown virtually no growth during the 1980s (chart). And its volatility has increased remark ably. Velocity growth in the 1980s (measured from the fourth quarter of one year to the fourth quarter of the next) has already ranged from - 5 .6 percent to +5.3 percent. Over the entire decade of the 1970s, the range was from -0 .1 percent to + 6.0 percent.1 Since the predictability of M1’s velocity is a key element in implementing a monetary targeting strategy, such dra matic changes in the behavior of velocity raise questions about what the underlying causes might be.2 This article explores some of the reasons for the changed behavior of M1’s velocity. The introduction of NOW accounts nationwide in 1981 is one factor. Another is the sharp decline in interest rates that has accom panied the reduction of inflation. In addition, swings in inventories and the deteriorating trade balance appear to be important. While the unusual behavior of velocity can be traced to several factors, these factors them selves, however, are not very predictable. Hence, movements in velocity measured in terms of GNP will probably continue to be difficult to anticipate. The first section of this article presents a brief review of recent movements in money, income, interest rates, and velocity. The second section analyzes the recent behavior of velocity using a conventional money demand equation. The final section presents an alternative analysis using the money-income reduced form equation.3 Review of recent velocity movements The declines in M1’s velocity in three of the last four years are certainly related to movements in interest rates (Table 1, column 3).4 In each year that velocity declined the Federal funds rate fell, with the largest decline in velocity occurring in the year with the largest percentage drop in the funds rate (1982-11 to 1983-11, shown in Table 1, columns 2 and 3). In contrast, over the period from 1983-11 to 1984-11 the funds rate rose and velocity increased as well. Clearly, fluctuations in interest rates explain a large part of the movements in velocity. These movements reflect the public’s changing demand for money as the level of interest rates and the opportunity cost of holding M1 balances change. However, too much weight might be assigned to changes in interest rates if GNP is not a good proxy for ’ Velocity is the ratio of GNP to M1. The behavior of velocity during the 1960s was quite similar to the 1970s. It grew about 3 percent per year, and stayed in a range of - 0 . 2 to 5.9 percent. Econom ists tend to look at the relationship between money and GNP, i.e ., velocity, from two different perspectives, the demand for money and the reduced form equation. In the demand for money, the public’s holdings of M1 balances are related to current and lagged values of interest rates and GNP The interest rate variable measures the cost of holding funds in M1 as opposed to investing them, while GNP measures the need for money for transactions purposes. In the reduced form equation, the growth of M1 is viewed as the primary determinant of aggregate demand. Hence, the growth of nominal GNP is related to current and lagged values of M1. Both of these approaches are useful in analyzing unusual movements in velocity. *For more background on the 1982-83 decline in velocity, see “Monetary Targeting and Velocity”, Conference Proceedings, Federal Reserve Bank of San Francisco (Decem ber 1983). 4The one-year periods run from the second quarter of one year to the second quarter of the next so that the first half of 1985 could be included. 16 FRBNY Quarterly Review/Autumn 1985 the volume of transactions that is important for money demand. That is, in each of the three periods when velocity declined during the 1980s, GNP growth slowed because of a decumulation in inventories or a reduction of net exports or both. These two components of GNP may not generate demand for money to nearly the same extent as the other components of GNP. Hence GNP growth during the periods when velocity declined could have been understating the increase in the quantity of transactions balances demanded. Velocity should measure the number of times per year a dollar of M1 is used for transactions purposes. GNP, however, is a measure of total production which can differ from total transactions for many reasons. For example, if consumers increase their transactions bal ances to purchase more goods, but firms choose to liquidate inventories rather than increase production, GNP is unchanged w hile M1 grows, and velocity declines. Likewise, if consumers increase their money balances to purchase more goods, but buy imports made attractive by a strong dollar, the money supply increases while GNP is constant, and velocity declines. Also, U.S. exports may affect the demand for money balances in foreign countries more than in the United States. Very little demand for M1 may be generated domestically by exports if inter-business transactions at the various stages of the production process result in relatively small balances in the checking accounts of business firms, compared with the balances consumers would keep to purchase the final product. Hence, if U.S. exports decline because of weak foreign demand, GNP falls while M1 demand remains relatively unaffected, and velocity weakens. In general, it might be better to look at gross domestic final demand (GNP less inventory investment and net exports) when assessing the trans actions demand for M1.5 Inventories and net exports appear related to the recent declines in velocity measured in terms of GNP (Table 1, column 7). Over the past year, for example, gross domestic final demand has been running about two percentage points above GNP, and in the first half of 1985 when the decline in velocity was particularly sharp, the divergence was 3.2 percentage points. In the two earlier periods when velocity was declining, GNP growth was also weaker than gross domestic final 5As long as the em p irica l analysis is done in a long-run context, the distinction b e tw ee n G N P and gross dom estic final d e m a n d w ould not be all that im portant. T heir long-run a v e ra g e growth rates have been about the sam e. H ow ever, during the 1980s net exports and inventories h ave had m uch larg er effects than in the past and, therefore, the distinction b etw een G N P and gross dom estic final dem a n d has b e c o m e m ore im portant for understanding the dem a n d for M 1. For exa m p le , the m ean absolute d iffe re n c e b etw een the growth rates of G N P and gross dom estic final dem and has been 2.7 p e rce n ta g e points in the 1 980s c o m p a re d with 2.1 p e rce n ta g e points in the 1970s and 1.8 p e rc e n ta g e points in the 1960s. demand. Since the transactions demand for M1 was stronger than GNP, velocity growth (measured in terms of GNP) appeared unusually weak. If no allowance was made for the effects of inventories and net exports, then too much weight might be given to interest rates in explaining movements in velocity. Changes in net exports and inventories have also been an important source of quarter-to-quarter volatility in velocity. Table 2 presents the ten largest deviations in M1’s velocity (measured in terms of GNP) from its trend growth rate over the past ten years in descending order. The third column shows the reduction of the deviations when velocity is computed with net exports and inventories excluded from GNP. In every case, the deviation of velocity from trend becomes smaller, with an average reduction of four percentage points. M1 Velocity P e rc e n t 2 0 x \\ v Growth ,^ 5 -1 0 G N P /M 1 7 .5 sx\s Levels .^ 70 S h a d e d a re a s re p re s e n t p e rio d s of re c e s s io n , as d e fin e d by th e N a tio n a l B ureau of E c o n o m ic R e s e a rc h . S o u rc e s : U .S. D e p a rtm e n t of C o m m e rc e and B o a rd of G o v e rn o rs of th e F e d e ra l R e s e rv e S ys te m . FRBNY Quarterly Review/Autumn 1985 17 Analysis using a demand for money equation An econometric model of the demand for M1 can also illustrate the effects of inventories and net exports. In the conventional transactions approach, real GNP and short-term nominal interest rates, currently and in past quarters, determine the volume of real M1 balances.6 In this article, the difference between real GNP and real gross domestic final demand (that is, the impact of net exports and inventories on GNP growth) is an additional explanatory variable used to capture the effect noted in the previous section. A few calculations will show the contribution of this variable in the money demand equation. Ignoring time lags and the impact of the interest rate variable, assume an income elasticity of 0.5. That would yield a relation ship: m = 0.5y, where m is the growth rate of real M1 and y is real GNP’s growth rate. If GNP increases 10 percent, real M1 increases 5 percent. Including the dif•ln the past, the most conventional specification related the log level of real M1 balances to the log levels of a short-term interest rate, real GNP, and lagged real M1 balances. For example, see Stephen M. Goldfeld, "The Demand for Money Revisited” , Brookings Papers on Economic Activity III (1973), pages 577-638, and "The Case of the Missing Money” , Brookings Papers on Economic Activity III (1976), pages 683-739. More recent research, however, suggests changes in logs, rather than log levels, would be a better way to specify the equation. See, for further detail, James S. Fackler and W. Douglas McMillin, “ Specification and Stability of the Goldfeld Money Demand Function” , Journal of Macroeconomics (Fall 1983), pages 437-459. In such equations, the coefficient on lagged money balances is quite small, suggesting that the lag from income and interest rates to money demand is short. To avoid constraining both GNP and the interest rate to the same implicit lag structure by using a lagged dependent variable, in this article the current and lagged values were incorporated directly in the regression. It appears to be an important distinction to make because the interest rate is insignificant in the current quarter, but significant lagged one quarter. GNP, on the other hand, is significant in the current quarter, but insignificant lagged one quarter. ference between real income and gross domestic final demand (yf) would result in the following equation, assuming the elasticities of y and yf are both 0.5: m = 0.5y - 0.5 (y - y f) In this case, a 10 percent increase in GNP due to a 10 percent increase in gross domestic final demand causes m to increase 5 percent as in the previous example. However, if yf increases 10 percent but y does not increase because inventories are run down, m will still increase 5 percent. In other words, the transactions demand for m will increase when the volume of trans actions increases, even if GNP (the level of gross domestic production) does not increase because of inventory rundowns or increased imports. The empirical results show that inventory investment and net exports are statistically important in a money demand equation (Table 3). The estimated coefficient for this variable is highly significant (at the 98.6 to 99.9 percent levels) in the three sample periods, thus improving the explanatory power of the equation about 20 percent. Moreover, the coefficient on the current quarter’s GNP becomes more significant with the addi tion of this variable. To show the importance of this additional variable for tracking the growth of M1 during the past few years, the regression equation was simulated after estimating the coefficients with and without the additional variable (Table 4). The simulation results are reported using coefficient estimates obtained from the 1971-80 and 1975-84 sample periods. The earlier sample period allows for an 18 quarter simulation period beyond the last year used for estimation. Alternatively, the 1975-84 sample period includes several quarters important for obtaining good coefficient estimates. Movements in M1, GNP, interest rates, inventories, and net exports were Table 1 Recent Velocity Movements In percent Change in the level of Federal funds rate (1) Percentage change in the level of Federal funds rate (2) Velocity growth (3) 1984-11 to 1985-11 .. - 2 .6 -2 5 -1 .4 + 7.3 1983-11 to 1984-11 .. + 1.8 + 20 + 3.7 + 7.5 Time period 18 M1 growth (4) Gross domestic final demand growth (6) Difference (7) = ( 6 ) - (5 ) + 5.8 + 7.6 + 1.8 + 11.6 + 10.9 - 0 .7 Nominal GNP growth (5) 1982-11 to 1983-11 ... - 5 .7 -3 9 - 4 .6 + 11.9 + 6.7 + 8.4 + 1.7 1981-11 to 1982-11 - 3 .3 -1 8 - 0 .2 + 5.1 + 4.9 + 5.7 + 0.8 FRBNY Quarterly Review/Autumn 1985 quite sharp in the 1980s. Moreover, financial innovation and deregulation have affected the demand for M1 since the mid-1970s, suggesting that earlier data might bias coefficient estimates. In the 1971-80 sample period, including the difference between GNP and gross domestic final demand in the equation causes the average absolute forecast error of the one-quarter growth rate of M1 to fall 1.4 percentage points, or almost one-third. This, of course, still leaves an average quarterly miss of three percentage points. In the second sample period, ending the estimation period in 1984 leaves only two quarters to test the model’s ability to track actual money growth beyond the estimation period. However, these two quarters are of particular interest because of the extremely sharp decline in velocity. Therefore, the objective of this exercise is to see whether an equation estimated through the early 1980s, when v e lo c ity grow th slow ed and its v a ria b ility increased, could track this most recent acceleration in M1 growth. The equation predicts 9 percent growth for the first half of 1985, while the actual growth is 10.4 percent. This relatively accurate forecast results from the larger estimated interest rate elasticity (in absolute value) in the later time period that occurs when earlier Table 2 Ten Largest Deviations in Velocity (Quarterly growth rates, from 1975 to 1985) In p e rc e n ta g e points at annual rates D ate Deviation In velocity growth from 1975 to 1985 average U sing G N P U sing less inventories GNP and net exports D iffe re n c e in absolute value 1981-1 ...... 16.0 8.3 7.7 1982-IV ... —13.7 - 8 .6 5.1 1982-1 ...... -1 0 .8 -6 .1 4.7 2.9 1978-11 .... 10.3 7.4 1980-111 ... - 8 .3 - 6 .5 1.8 1985-11 .... - 7 .6 - 3 .6 4.0 1981-111 ... 7.5 4.7 2.8 1985-1 ...... - 7 .0 - 5 .8 1.2 1975-111 ... 6.8 3.1 3.7 1984-1 5.8 0.8 5.0 M ean absolute av e ra g e 9.4 5.5 3.9 data are excluded and from the additional variable to control for the effects of inventories and net exports.7 Analysis using a reduced form equation Another way to analyze velocity movements is by using a reduced form equation relating the current quarter’s GNP growth rate to current and past M1 growth.8 In this section, the analysis with the reduced form equation shows that much of the apparent instability in velocity, particularly in 1982 and 1985, stems from inventories and net exports as well as from the introduction of nationwide NOW accounts in 1981. The reduced form equation says that GNP growth equals average velocity growth plus a weighted average of M1 growth in the current and four past periods. In other words, recent M1 growth is the primary determi nant of current nominal aggregate demand. The basic form of this equation is shown as equation 1 in the right side of Table 5. To further refine this relationship, an article in an earlier Quarterly Review showed that M1 growth coming from other checkable deposits (OCD) tends to have only a little more than half of the impact on GNP that M1 growth coming from currency and demand deposits (MA) has.9 This result appears in equation 2. The third equation in Table 5 is the same as the second equation except that gross domestic final demand (YF) replaces GNP (Y) as the dependent variable. In the context of the reduced form equation, the logic for subtracting inventories and net exports from GNP is different from that for money demand. In this case, stronger M1 growth creates greater demand for goods and services, but if imports or inventories satisfy some of that demand, GNP growth does not pick up as much 7O ther analysts have noted that the interest elasticity in the conventional m oney d e m a n d equation in creases in a bsolute value when the sam ple period exc lu d e s earlie r d a ta . In part, this could be due to the nationw ide introduction of N O W a c counts in 19 8 1. N O W accounts e arn e xplicit interest and consum ers with N O W accounts could be m ore sensitive to c h a n g e s in m arket rates than those with dem and deposits. M oreover, with the introduction of m oney m arket funds and M M DAs, it has b e c o m e e a s ie r for consum ers to shift their liquid assets into and out of M1 w hen m arket rates c h a n g e . For more detail, see H ow ard Roth, “ Effects of Financial D e re g u la tio n on M onetary Policy", Economic Review, Federal R es erv e Bank of K ansas C ity (M arch 1985); and M .A . Akhtar, “ Financial Innovations and Their Im plications for M onetary Policy: An International P erspective", Bank for International Settlem ents, Economic Papers No. 9 (D e c e m b e r 1983). •O ve r the years, m any objections have been raised to the red u c ed form app ro a c h . In particular, M 1, like GNP, is an end o g e n o u s v ariable and the correlation o b s e rve d in the red u c ed form equation results from both v a ria b le s respo nding in a system atic w ay to other factors in the econom y. Even if M1 is not exogenously d e te rm in e d , however, this relationship can be useful if M1 respo nds sooner to these other factors and hence is a good lead in g in d icato r of GNP. For m ore detail, see John W enninger, “The M 1 -G N P R elationship: A C om ponent A pproach", this Quarterly Review (A utum n 19 8 4). •W enninger, op. cit. FRBNY Quarterly Review/Autumn 1985 19 Table 3 Estimation Results for the Demand for Money Sample period Dependent variable Coefficient estimates Summary statistics Constant r r( —1) y y( - 1 ) N E+II P R2 SE 1960-84 ................ m -0.0 0 1 3 (10 ) 0.0065 (1.1) -0 .0 3 9 (6.3) 0.32 (4.1) 0.094 (1.2) * 0.28 0.41 0.0070 1960-84 ................ m -0.0024 (1.9) 0.0063 (1.1) -0.031 (5.3) 0.49 (5.8) . 0.036 (0.5) —0.49 (4.1) 0.29 0.48 0.0065 1971-80 ................ m -0.0 0 4 3 (2.7) 0.014 (16) -0 .0 2 7 (3-2) 0.48 (4.3) -0 .0 1 3 (0.1) * 0.10 0.51 0.0070 1971-80 ................ m -0.0 0 5 6 (3.8) 0.011 (13) -0.021 (2.6) 0.63 (5.3) -0 .0 0 9 (0.1) -0 .5 5 (2.5) 0.06 0.58 0.0065 1975-84 ................ m -0.0001 (0.0) 0.0055 (0.5) -0.061 (5.8) 0.24 (1.8) 0.095 (0.7) * 0.32 0.50 0.0081 1975-84 ................ m -0.0014 (06) 0.0038 (0.4) -0 .0 4 5 (4.7) 0.46 (3.6) -0 .0 0 5 (0.0) -0 .6 9 (3.9) 0.38 0.61 0.0069 Definition of variables: m = Ain (M1/GNP deflator), r = Ain (3-month Treasury bill rate), y = Ain (real GNP). NE + II = A(ln [real GNP] - In [gross domestic final sales/GNP deflator]). 'Not included. Table 4 Simulation Results for the Demand for Money In percent at annual rates Predicted M1 growth using equation estimated 1971-80 Actual M1 growth Without net exports plus inventory investment With net exports plus inventory investment Without net exports plus inventory investment With net exports plus inventory investment 1981-1......................................... 1981-11 ........................................ 1981-111 ...................................... 1981 -IV ...................................... 3.3 8.8 3.1 5.1 6.6 - 2 .2 14.0 - 3 .8 5.1 - 0 .2 12.3 - 2 .4 0.7 5.5 12.4 - 0 .2 - 0 .2 6.1 11.5 0.6 1982-1 ......................................... 1982-11 ........................................ 1982-111 ...................................... 1982-IV ...................................... 8.9 2.9 5.9 16.3 6.9 7.3 0.6 10.2 7.8 4.9 4.0 10.4 8.1 2.3 3.5 12.9 9.7 0.3 6.9 12.6 1983-1 ......................................... 1983-11 ........................................ 1983-111 ...................................... 1983-IV ...................................... 11.3 12.2 10.2 6.3 19.0 9.7 11.5 9.9 13.9 13.5 10.6 10.1 16.1 6.0 12.4 9.6 10.1 11.3 10.7 10.2 1984-1 ......................................... 1984-11 ........................................ 1984-111 ...................................... 1984-IV ...................................... 6.2 6.5 4.5 3.2 10.1 3.1 4.2 3.6 8.0 7.2 2.2 3.6 10.4 3.0 4.9 3.3 7.7 8.1 2.6 3.7 1985-1 ......................................... 1985-11 ........................................ 10.6 10.2 6.7 7.7 6.3 9.8 10.5 5.7 9.0 9.1 4.5 3.1 3.0 2.3 Date Average absolute e rr o r ........... 20 Predicted M1 growth using equation estimated 1975-84 FRBNY Quarterly Review/Autumn 1985 average absolute error (lower half of Table 5). The average absolute error declines from 5.3 to 4.4 per centage points when OCD and MA are allowed to have different-sized impacts, and declines further to 2.7 percent when gross domestic final demand is used as the dependent variable. The reduction of the error for the period as a whole is found in mostly 1981, 1982, and the first half of 1985. The questions remain whether GNP growth in indi vidual quarters has been particularly difficult for these equations to track and whether the distinction between GNP and gross domestic final demand would have made any difference in those quarters.10 Table 6 shows as would be expected. Slow GNP growth relative to M1 growth reduces velocity from what it would have been if domestic production had risen. Likewise, the demand for exports can weaken significantly for reasons unre lated to M1 growth; for example, sluggish growth in the economies of our trading partners. Reduced demand for exp orts w eakens GNP but leaves M1 grow th unchanged, causing velocity growth to slow. The left side of Table 5 shows the simulation errors from each of these three equations. Average errors appear in the upper half of the table and average absolute errors in the lower half. The average error (a measure of bias) for the entire period falls from - 2 .8 percentage points to -1 .1 percentage points when OCD and MA are allowed to have different impacts on GNP growth. It declines further, to just - 0 .4 percentage point, when YF replaces Y as the dependent variable. The reduction of the average error for the entire period stems mostly from better performance in 1982 and in the first half of 1985. Another striking improvement is the decline in the 10One way of exploring this question is to include a zero-one dummy variable for each quarter since 1979. Those dummy variables that are statistically significant—the estimated coefficient before the dummy variable is significantly different from zero using a t-test— occur in quarters where the equation had significant forecast errors For more on this approach, see R.W. Hafer, “ Monetary Stabilization Policy: Evidence from Money Demand Forecasts” , Federal Reserve Bank of St. Louis Review (May 1985). Table 5 Reduced Form Results In percentage points at annual rates In-sample average errors Y on M (1) 1980 ............................ -0 .1 Y on MA, OCD (2) YF on MA, OCD 0.0 0.1 1981 ............................ 1.5 4.1 2.7 1982 ............................ - 7 .6 - 4 .2 - 1 .8 1983 ............................ - 3 .6 - 1 .9 -3 .4 1984 ............................ 0.6 2.0 1.6 1985 (first half) ......... - 6 .3 - 5 .8 - 2 .6 1980-85 ........................ - 2 .8 -1 .1 -0 .4 (1) Y = 3.4 + 0.97M (6.2) 0.23 (2) Y = 2.9 + 1.17MA + 0.65 OCD (6.7) (3.4) 0.27 (3) YF = 3.4 + 1 08MA + 0.660CD (7.1) (3.9) 0.30 Sample periods: 1949-11 to 1985-11 Y = quarterly growth rate of GNP. In-sample average absolute errors 1980 ............................ 2.4 2.5 2.7 1981 ............................ 6.8 6.2 3.3 1982 ............................ 7.6 4.2 1.8 1983 ............................ 3.6 2.6 3.5 1984 ............................ 2.3 2.9 2.1 1985 (first half) 6.3 5.8 2.6 1980-85 ........................ 5.3 4.4 2.7 _ R2 Equations M = quarterly growth rate of M1. OCD = quarterly M1 growth due to the other checkable deposit components of M1. MA = quarterly M1 growth due to M1 less OCD. YF = quarterly growth rate of GNP less inventories and net exports. The equations are estimated with polynominal distributed lags covering the current quarter and four lags. FRBNY Quarterly Review/Autumn 1985 21 Conclusions While it is not possible to account precisely for every quarterly movement in velocity, several factors have played important roles in recent years. From the point of view of money demand, these factors include the declines in interest rates, an increased responsiveness in the public’s demand for M1 when interest rates change, and the consideration that GNP is not a good proxy for the total volume of transactions when net exports or inventories are strongly affecting its growth rate. From the perspective of the reduced form equation, the errors in predicting GNP with M1 are lowered when M1 growth is split into its interest bearing and non interest bearing components, and when the distinction be tween GNP and gross domestic final demand is made. However, it is very difficult to predict swings in inventories, net exports, interest rates, and the split in M1 growth among its components. Moreover, there has not been enough experience with M1 in this more de regulated environment to estimate very precisely the interest elasticity of the demand for M1. Hence, even though some of the reasons for the instability of velocity in the 1980s (measured in terms of GNP) can be iden tified ex post, velocity is not likely to be more predict able as a result. the results by year for GNP and gross domestic final demand.11 In terms of GNP, four quarters out of 22 in the sim ulation period show statistically significant errors ranging from 10.7 to 13.5 percentage points: 1981-1, 1982-1, 1982-IV, and 1985-11. In all four cases, however, the errors become smaller (roughly half as large) in absolute value and turn statistically insignificant when gross domestic demand rather than GNP is used as the dependent variable. But the error in the first quarter of 1983 becomes larger in absolute value and turns sig nificant when gross domestic final demand is used. In that quarter, when net exports and inventories were adding five percentage points to GNP growth, its growth was still considerably weaker than would have been expected from the very rapid pace of M1 growth. Hence, it appears that some “ outliers” will still occur from time to time, even though the distinction between gross domestic final demand and GNP can reduce many of the large errors in the reduced form equation. 11The distinction between OCD and MA could not be made in this exercise. Nationwide NOWs were introduced in 1981. With a dummy variable for each quarter in the post-1979 period, it is not possible for the regression to assign separate weights to OCD and MA. Table 6 Significant Errors in Reduced Form Equations In p e rc e n ta g e points at annual rates Q uarter 1 980 1982 1981 Y YF Y YF Y YF Y ....................... 1.8 (0 .4 ) 1.0 (0 .2 ) 10.7 (2 .3 )* 3.1 (0 .7 ) -1 1 .8 (2 5 )* - 5 .9 (1 .4 ) — 9.4 (1 .9 ) II ............................... -1 .3 (0 .2 ) -5 .0 (1 .1 ) -6 .1 (1 .3 ) -4 .5 (1 .1 ) -3 .8 (0 .8 ) -5 9 (1 .4 ) Ill ............................... -3 .8 (0 .8 ) -0 .5 (0 .1 ) 5 .3 (1 .1 ) 2 .0 (0 .5 ) -6 .8 (1 .5 ) I V ............................... -1 .0 (0 .2 ) 1.0 (1 .2 ) -5 .7 (1 2 ) -4 .1 (1 0 ) A v erag e error ... A v erag e a bsolute error .. . -1 .1 -0 .9 1.1 2 .0 1.9 7.0 I 1984 1983 YF 198 5 Y YF Y YF -1 2 .5 (2 .9 )* 4 .6 (1 0 ) -0 .9 (0 .2 ) -6 .5 (1 .4 ) -4 .4 (1 .0 ) -6 .0 (1 .2 ) -4 .7 (1 1 ) 0 .7 (0 .1 ) 4.2 (1 .0 ) -1 0 .0 (2 .1 )* -4 .3 (1 .0 ) -2 .2 (0 .5 ) -5 .9 (1 .2 ) -6 .3 (1 5 ) -3 .2 (0 .6 ) -2 .3 (0 .5 ) -1 3 .5 (2 8 )* -5 .9 (1 -3 ) -1 .3 (0 .3 ) -2 .3 (0 .5 ) -0 .2 (0 .0 ) -0 .6 (0 .1 ) -0 .9 -9 .0 -5 .0 -5 .7 -6 .5 0 .5 0.1 -8 .3 -4 .3 3.4 9.0 5.0 5.7 6.5 2 .2 2 .0 8 .3 4 .3 Equations: Y = 2 .9 + 1 .1 9 M + dum m y v a ria b le for e ach p o s t-1 9 7 9 quarter. (6 .7 ) YF = 3 .2 + 1 .1 2 M + dum m y v a ria b le for e ach p o s t-1 9 7 9 quarter. (6 7) 'S ig n ific a n t at 9 5 p e rce n t level, see notes in Table 5 for exp lan atio n of variables. Lawrence J. Radecki and John Wenninger 22 FRBNY Quarterly Review/Autumn 1985 The Strong Dollar and U.S Inflation U.S. inflation has changed remarkably little during the present recovery. Consumer prices rose at a 3.8 percent annual rate during the first half of 1985, barely different from the 3.7 percent increase posted for the first year of expansion. The steadiness of the inflation rate over the past two and one-half years is somewhat surprising in view of several factors that might have reduced it further. Oil and several other key commodity prices have fallen sharply since 1982 (Table 1), while significant slack remains in labor markets, as indicated by an unem ployment rate still above (according to most analysts) the “full-employment” level. In addition, the dollar has appreciated nearly 17 percent (trade-weighted average basis) over the same period (Chart 1). In the past, these conditions have often been associated with falling inflation—so why not during this recovery? This article focuses on the dollar’s impact on U.S. inflation over the last several years. The dollar’s rise since 1982 has not led to the fall in aggregate import prices that past experience would have suggested, perhaps helping to explain why inflation has not mod erated further. Much of the surprising relative strength of import prices can be attributed to the sharp recovery in domestic real growth, which led to increases in import demand that substantially offset the downward pressure on import prices from the dollar appreciation. This experience suggests that the dollar depreciation since February may not add much if at all to domestic inflation unless domestic demand picks up markedly from the sluggish pace of the first half of 1985. Experience The recent pattern of a strong dollar with virtually unchanged domestic inflation differs considerably from 1980 to 1982, when the dollar rose by 20 percent while the inflation rate fell nearly eight percentage points. It differs as well from the late 1970s experience of dollar depreciation accompanied by rising inflation. Of course, other factors, notably substantial differences in gov ernment policies, were primarily responsible for this contrast. Still, the impression persists that inflation has not responded to the dollar as much in the last two years as it did in the past. Statistical estimates of the response of domestic prices to changes in the dollar, most derived from data drawn largely from the 1970s, reinforce this impression. Though estimates vary substantially, depending on the model and period of estimation (appendix), the con sensus is that a 10 percent rise in the dollar’s value will reduce the Consumer Price Index (CPI) inflation rate by about 0.6 percent in each of the following two years. On this basis, the dollar’s appreciation since the last cyclical trough should have reduced the CPI by nearly 1.5 percent below the level it would otherwise have reached. But such a dampening effect on inflation from the rising dollar is not obvious from the actual data.1 This raises a natural question prompted by the sub stantial fall in the dollar since last February: will U.S. inflation remain unaffected, or will it rise as the expe rience prior to 1982 might suggest? Import prices That the relation between exchange rate movements and inflation seems to vary is not surprising since the two are linked through several channels.2 Changes in ’This is not to say that inflation did not fall through 1984. Rather, the extent of that decline, 0.4 percentage point, was slight relative to the movements in factors generally thought to influence inflation. 2By “linkage" we mean an association between the two endogenous variables (exchange rates and prices), not a statement about causation. FRBNY Quarterly Review/Autumn 1985 23 Chart 1 Inflation and the Value of the Dollar Percent Index 1980=100 Year-over-year annual rate. ^Trade-w eighted average of dollar's value vis a vis 12 industrial countries’ currencies; weights are bilateral shares of U.S. trade. Chart 2 Price and Dollar Cost of Imports U.S. dollar index, 1980=100 110--------------------------------------------------------------------------------------------------------------- Non-oil im port p r ic e * the dollar directly affect import prices, which are com ponents of the CPI and other dom estic price level measures. Dollar appreciation, for example, reduces the cost expressed in dollars of foreign produced goods, allowing import prices to fall without any reduction of foreign exporters’ profit margins. However, the extent to which this cost-reduction is “ passed-through” to import prices may change with economic circumstances. Fur thermore, the response of inflation to the dollar will also depend on how domestic product prices and wages are affected by import price changes, on the response of government policies, and possibly on other factors as well. Thus there are several potential explanations for the apparent change in the relation between the dollar and-U.S. inflation in recent years. Nonetheless, the following data suggest that a shift in the pass-through of dollar cost changes to import prices may be a significant part of the explanation. The trade-weighted value of the dollar increased by nearly 17 percent from the first quarter of 1983 through the second quarter of 1985, while foreign export costs (as measured by local currency export prices) rose by an average of 8 percent. Taken together, these suggest that the cost expressed in dollars (“ dollar cost” ) of goods exported to the United States has declined by over 7 percent since the first quarter of 1983.3 Since aggregate non-petroleum import prices have risen by nearly 0.4 percent over the same period, there effectively has been no pass-through of this change in dollar cost to the average price paid for imports in the United States. (Pass-through, as defined here, is the ratio of the actual change in import price to the change in dollar import cost over a given period). In effect, foreign exporters’ profit margins have widened significantly with dollar apprecia tion. Note, however, that while nominal import prices have remained nearly flat, they have fallen substantially relative to prices of domestically produced goods.4 3The dollar cost of im ports refers to their foreign p roduction cost (in local c urrency) c o nverted to dollars at p revailing e x c h a n g e rates. Thus, for exam p le, a 10 p e rce n t rise in the dollar w ould, all other factors u n c hanged, lower the dollar cost of U .S . im ports by the sam e am ount. U sing a g g re g a te foreign export price indexes to m easure the local c u rren cy production cost c le arly is only a p proxim ate (in part b e c a u s e the com position of a g g re g a te foreign exports m ay differ from that of their exports to the U n ited S tates). H ow ever som e a lte rn ative m easures (e .g ., foreign w h o les a le p rices) lead to very sim ilar conclusions. *N o n -o il im port unit value index. + Measured as the average of foreign e xp o rt p rice indexes of 12 countries converted to dollars; weights are bilateral shares of U.S. trade. Sources: International Monetary Fund, International Financial S ta tistics, various years and U.S. Department of Commerce. 24 FRBNY Quarterly Review/Autumn 1985 4lm port prices have d e c lin e d nearly 9 p e rce n t relative to the CPI since the first quarter of 1 9 8 3, so that do llar a p p re c ia tio n has had a s ignificant im pact on the "re a l" price {i.e., relative to pric es of dom estic substitutes) and volum es of im ports. Furtherm ore, the virtually zero pass-through (as d e fin e d h e re) d o es not m ean that im port prices necessarily w ould h ave rem ain e d u n c h a n g e d had the dollar not a p p re c ia te d . In d e e d , the arg um ents later in the text and in the box suggest that im port pric es w ould h ave risen significantly further had the dollar s tayed at its first q u a rte r 1 9 8 3 level. The pass-through over the current recovery has been substantially lower than that seen in 1981-82, and strikingly lower than during the late 1970s (Table 2 and Chart 2). Indeed, the pass-through was more than complete over 1977-78, when the dollar was depre ciating and U.S. inflation was rising, while it was about one-quarter over 1981-82, when inflation was declining. Underlying the apparently low pass-through of the dollar appreciation to aggregate import prices is a fairly wide divergence among major product components (Table 3). The average price of imported automobiles (including parts) has increased nearly 10 percent since the cyclical trough, and over 30 percent since the end of 1980. Prices of imported capital goods have also risen over the recovery while imported consumer goods’ prices have fallen only slightly. The price of industrial supplies (and of agricultural imports since 1980) has, by contrast, fallen considerably more. This divergence also differs from the 1977-78 period, when, except for autos, the increase in prices was significantly more uniform among categories. The rise in auto prices after 1980 might be considered a special factor that has distorted the measured pass through. This is because imports from Japan (which account for the bulk of total imports of finished autos) until recently were lim ited by an effective quota. Because of this quota, the dollar’s rise is unlikely to have affected auto import prices significantly over this period. The price of imports excluding autos and parts has fallen by nearly 2.5 percent during the recovery, and by nearly 9 percent since the end of 1980, but the implied pass-through is still well below that for 1977-78. Possible explanation The apparently low pass-through of the dollar’s appre ciation to import prices might seem to reflect “ monop olistic” or other noncompetitive practices. However, there is an alternative explanation that seems reason ably consistent with the actual record and is compatible with competitive behavior by exporting and importing firm s.5 This is based on changing relations among inflation, growth, and exchange rates since the 1970s, which have altered movements of import costs relative to the domestic demand for imports. The dollar’s depreciation over 1977-78 was substan tially offset by differential U.S.-foreign inflation. Con sequently, the dollar cost of imports from abroad, U.S. import prices, and the prices of domestically-produced goods all rose togethe r and by roughly the same amount. By contrast, the dollar’s rise since 1980 has *This explanation is not meant to exclude the possibility of oligopolistic or monopolistic practices, at least in some industries. Furthermore, it does generally presume that U.S. import demand is a significant share of the world total. Table 1 Consumer Prices and the Exchange Rate Percent change End of period level: Unemployment rate (4) Period United States CPI (1) 1985-11/1984-11 . 1983-IV/1982-IV 1982-IV/1981 -IV 1981-IV/1980-IV 1980-IV/1979-IV .. 3.7 .. 3.3 .. 4.5 .. 9.6 .. 12.5 9.5 0.8 11.1 9.4 - 0 .7 - 1 .9 -1 3 .7 - 5 .0 8.6 43.2 7.3 8.5 10.6 8.2 7.4 1985-11/1983-I .. .. 9.3 1982-IV/1980-IV .. 14.4 1978-IV/1976-IV .. 16.2 16.5 21.5 - 9 .9 -1 3 .4 3.2 9.8 7.3 10.6 5.9 Dollar Index exchange of oil rate* prices (2) (3) ’ Trade-weighted average value of the dollar vis-d-vis currencies of 12 foreign industrial countries. Table 2 Import Prices and the Exchange Rate Percent change Foreign Dollar export exchange cost* rate (2) (1) Period 1985-11/1983-I 1982-IV/1980-IV . 1978-IV/1976-IV 8.0 9.8 7.9 16.5 21.5 - 9 .9 Dollar import co stf (3) - 7 .3 - 9 .6 19.8 Import price (4) 0.4 - 2 .3 23.8 •Foreign export cost is measured as a trade-weighted average of export prices (in local currency) of 12 foreign industrial countries. fChange in foreign export cost.expressed in dollars (approximately equal to column 1 minus column 2). Table 3 Components of Import Price Change Category Percent change unit value over: 1985-11/ 1982-IV/ 1978-1V/ 1983-1 1980-IV 1976-IV Total non-oil .............................. Autos ......................................... Capita! ....................................... Consumer ................................. Industrial supplies ................... Food, feeds, and beverages 0.4 9.7 7.3 - 3 .3 - 8 .2 - 0 .3 - 2 .3 17.5 - 6 .9 3.2 - 7 .6 -1 3 .6 23.8 34.7 22.8 19.8 16.0 16.7 Import price excluding autos.. - 2 .4 - 5 .3 21.2 FRBNY Quarterly Review/Autumn 1985 25 greatly exceeded U.S. relative to foreign inflation.6 Thus the dollar cost of imports has fallen during the 1980s while domestic U.S. prices have continued to rise, although more slowly than before. In short, cost pres sures reinforced domestic demand pressures to push im port prices up during the late 1970s, but more recently these forces have tended to offset one another. ®This amounts to saying that the dollar’s real value— its nominal value adjusted for U.S.-foreign inflation—has risen sharply since 1980, whereas it changed considerably less over 1977-78. Supply and Demand Explanation The argument can be put in the familiar supply and demand framework. The supply of imports typically increases with the ratio of the domestic selling price to the dollar cost of their production. This is represented by the upward-sloped supply schedule in Chart 3. Import supply also increases with foreign export capacity (which shifts the supply curve). Import demand declines as the domestic import price rises relative to the prices of d o m estically produced products, as shown by the downward-sloped schedule in Chart 3, and increases with domestic real income. An exchange rate depreciation amounts to a reduction of supply—an upward shift in the supply schedule. With no change in demand, the extent of pass-through depends on the relative slopes of import supply and demand, and will generally be incomplete. The pass through will be greater the more elastic is supply and inelastic is demand, and will be complete only if supply is perfectly elastic or demand inelastic. (More generally, the pass-through from an exchange rate change, given no change in domestic or foreign prices and incomes, is equal to the ratio of the supply price elasticity to the sum of the supply and demand price elasticities.) However, when exchange rate depreciation is accom panied by domestic price and income increases, the demand schedule also shifts up (Chart 4). In this case, domestic demand increases reinforce the impact of dollar depreciation in raising the dollar cost of imports (the shift in supply), leading to a higher pass-through than when supply alone is shifting. The observed response of imports to the depreciation will thus be greater the more demand increases. Indeed, if domestic prices increase (relative to abroad) by the same proportionate amount as the exchange rate depreciates, the observed pass through will be complete, regardless of the elasticities of import supply and demand (unless real growth rates diverge considerably). This is essentially the environment that prevailed over 1977-78, during which the pass through of the dollar’s decline appeared virtually com plete in nearly all major import categories. 26 FRBNY Quarterly Review/Autumn 1985 This can be seen in terms of the specific contributions of changes in import costs and import demand to import prices. To a foreign supplier sending goods into the U.S. market, a dollar depreciation amounts effectively to a proportionate increase in the dollar cost of delivering a given amount. But the extent to which this increase in cost is passed-through to the actual dollar import price also depends on what is happening to import demand. If demand is not growing, the foreign supplier can fully pass-through the increased cost to the price only by selling less than before. For this reason, the price is apt to rise somewhat less than the cost, that is, the pass through will be less than complete, and exporters’ profit margins probably will fall. However, the pass-through is apt to be greater if demand is increasing, either because prices of dom estically produced goods are rising, making imports more attractive, or because domestic real income is growing. More generally, this implies that the apparent impact of exchange rate changes on domestic import prices is likely to be significantly greater when cost and demand pressures are reinforcing one another than when they are not (box). In the general inflationary environment of 1977-78, the increasing dollar cost of im ports associated with exchange depreciation was accompanied by increasing domestic prices and real income and hence increasing demand for imports. The apparent pass-through would be expected to be relatively high under these circum stances. This is because the effect of rising domestic prices and income on domestic demand for imports reinforced the exchange rate depreciation in pushing up import prices. Furthermore, with costs and demand pressures moving so closely together, it is not surprising that the pass-through was virtually complete— and in all major categories. Since 1980, however, the dollar’s appreciation has led to a fairly steady decline in import dollar costs. A sig nificant portion of this cost decrease continued to be passed-through to prices over 1981-82, in large part because domestic demand growth also slowed mark edly.7 Subsequently, however, aggregate demand has grown fairly rapidly on average, so that the falling dollar cost of imports has been partially offset by the upward pressures on import demand from rising domestic prices and strong real income growth. This may largely explain why the pass-through of exchange rate changes to import prices now appears to be much lower than before (as well as why exporters’ profit margins have widened). And with import prices varying with exchange rates less 7Pass-through averaged 25 percent over 1980-IV to 1982-IV although there was considerable variation within the interval. Despite the dollar’s appreciation, substantial pass-through would be expected during this period given that weakening domestic activity probably exerted little, if any, offsetting pressure on import prices. than before, the seeming failure of inflation to respond to recent dollar appreciation is more understandable. The same patterns emerge in the data for major product categories. Investment and, to a lesser extent, consumer spending have been unusually strong (on average) over the current recovery, suggesting that the domestic demand influence on the prices of these products has been expanding relatively rapidly. These factors may help explain why the dollar’s appreciation seems to have had especially little impact on prices of imported capital and consumer goods.8 Chart 3 Determination of Import Price Under Depreciation Price Chart 4 Depreciation Accompanied by Domestic Price Inflation Price (PM) Import supply ( d e p re c ia tio n ) Import dem and (dom estic inflation) If the dollar falls If the changing import pass-through over the past ten years is due to shifting import cost and domestic demand influences, then it could be m isleading to extrapolate mechanically from recent behavior to assess the implications of future dollar movements. Suppose, for example, that the dollar were to fall substantially from its present level over the coming year. What would be the likely impact on inflation of this reversal? The pass-through observed during the last two and one-half years might suggest no significant change in either import prices or domestic consumer prices. However, this overlooks a fundamental change in import cost relative to demand movements that could occur with a dollar decline. A substantial dollar depreciation would raise the dollar cost of imports considerably, reversing the pattern of the last two and one-half years. The response of import prices again will largely depend on the course of domestic demand. If strong U.S. growth were to resume, the cost and demand influences would reinforce one another, leading to a higher pass-through than has been observed over the last several years. Indeed, prior experience suggests that the pass-through to import prices could be as high as 50 to 70 percent. This means that exporters’ profit margins would absorb one-third to one-half of a sub stantial dollar depreciation, or perhaps even more given that these margins are now relatively high, with the remainder passed on to higher import prices. On the other hand, if the economy were to expand sluggishly, the demand pressures on import prices would be much less, even absent. In that case, shrinking profit margins probably would largely offset a dollar decline, leaving little if any impact on import prices. 8As explained in the box, the pass-through would be expected to be lower the more elastic is demand. Previous studies suggest that demand for imported capital and consumer goods is more price elastic than that for materials and agricultural imports, which also helps explain the contrast. Quantity Charles Pigott and Vincent Reinhart FRBNY Quarterly Review/Autumn 1985 27 Some Recent Evidence on the Dollar/Price Link This appendix surveys the recent literature on the impact of dollar depreciation on domestic inflation. These results are summarized in the table. Two main strategies have emerged in the work on the inflation consequences of exchange rate changes. In the first, a “ small-model" approach, an import price variable is included among the explanatory variables in a standard inflation-determination equation. This is the approach taken in the small models summarized in the first portion of the table. To estimate the impact of exchange rates on domestic prices, we must first gauge th e ir im pact on im port prices. In what follow s we assumed that 60 percent of a change in the exchange rate is passed-through to import prices.* The second approach to judging the impact of the dollar appreciation is to consider the predictions gen erated by large scale macroeconometric models where the linkages between exchange rates and dom estic prices are made explicit in a number of equations. Such structural models often report the impact on both con sumer prices and the GNP deflator, and typically it is the former that increases more. This is because imports directly enter the CPI but enter the deflator only indirectly (through the prices of domestically produced goods). *To replicate Gordon’s results his basic intlation determination equation was re-estimated. The coefficient estimates obtained, which are close to those Gordon reports, are used for the simulation results reported in the table. The Impact on Domestic Inflation of a 10 Percent Dollar Depreciation in One Quarter Measured as percentage points added to average yearly rates Study Price index First year Second year Remarks Small model results Dornbusch-Krugman (1976) ...... Consumer prices 0.8 0.5 • Import prices are included in a standard inflation determination equation. • Elasticity of CPI inflation with respect to import price inflation is 0.14 in the short run, 0.42 in the long run. •Estimates use annual data from 1957 to 1973. •We assumed a pass-through of 0.6. Kwack (1977) ................................ Consumer prices 1.5 0.3 *The model specifies the price linkages for 12 coun tries, determining consumer, import, and export prices. •Estimates from 1957 to 1973 use annual data. •A 1 percent change in the exchange rate causes a more-than-complete pass-through of 1.23. •We simulated the U.S. sector in isolation. Spitaller (1978) ............................ Consumer prices 0.5 0.5 •Estimates are derived from CPI inflation equation using money growth, industrial production relative to trend, and import price inflation. • Elasticity of CPI inflation with respect to import price inflation is 0.04 in the short run and 0.27 in the long run. •Estimates from 1958 to 1976 use four-quarter rates of change. •We appended a pass-through equation. Fixed weight GNP deflator 1.1 0.8 *The model estim ates an in fla tio n d e te rm in a tio n equation using la g g e d in fla tio n , exch a ng e rate changes, the unemployment rate, and dummy vari ables. • Gordon does not report enough coefficients to simu late the model so we re-estimated over the quarterly data from 1975 to 1984. Gordon (1982 and 1983) FRBNY Quarterly Review/Autumn 1985 Some Recent Evidence on the Dollar/Price Link, continued The Impact on Domestic Inflation of a 10 Percent Dollar Depreciation in One Quarter Measured as percentage points added to average yearly rates Study Price index First year Second year Remarks Large model results Federal Reserve Board of Governors’ Multi-Country Model Consumer prices IMF's Multilateral Exchange Rate Model .................................... Consumer prices Low feedback ........................... High feedback ........................... OECD interlink Federal Reserve Board of Governors’ FMP Model .... Data Resources 0.5 0.5 1.4 4.4 (total impact) • The model links domestic macro models for the United States, Germany, Japan, the United Kingdom, and the rest of the world. •The equations, with a few exceptions, were estimated over the quarterly observations available from 1961 to 1975. •This is a mathematical simulation model with a com plete microeconomic specification in which a priori judgment is used in the choice of parameters. »The MERM estimates the medium term (two to three years) effects of exchange rate changes. •The low feedback case assumes that a 1 percent increase in the CPI raises wages by 0.5 percent. •The high feedback case assumes that a 1 percent increase in the CPI raises wages by 0.85 percent. Domestic demand deflator 1.0 0.4 • The model groups together m edium -sized macro models (about 150 equations each) for 23 countries. • Some of the coefficients are estimated with the rest assigned according to the judgment of the modelers. Consumer prices GNP deflator 0.8 0.5 0.5 0.3 • This is a quarterly model with approxim ately 500 equations. • There is a complete modeling of capital flows, with exchange rates endogenous to the system. Consumer prices GNP deflator 0.4 0.1 0.3 0.4 "This is a quarterly model with approximately 1200 equations. • The exchange rate is determined endogenously. FRBNY Quarterly Review/Autumn 1985 29 Federal Deposit Insurance and Deposits at Foreign Branches of U.S. Banks Should the Federal Deposit Insurance Corporation (FDIC) charge insurance premiums on deposits in for eign branches and International Banking Facilities (IBFs) of U.S. banks? Such a proposal has appeared as one of many possible changes to the Federal deposit insurance system, but the issue has received relatively little attention.1 This article airs the issues involved in an extension of the FDIC premium to foreign branches without taking a position on the question. Levying premiums on these deposits would alter the distribution of premium charges significantly. But as this study shows, how equitable the proposed redistribution would be depends on how one views key characteristics of FDIC insurance coverage. Further, the change could have important repercussions for the competitive structure of banking inside and out side the United States. The nature of the proposal Several proposals have been made to include deposits at foreign branches of U.S. banks in the base used to compute FDIC insurance premiums. These proposals The author would like to thank Edward Frydl, Sherrill Shaffer, Robert McCauley, and Melissa Berman for their comments, and David Bush for his assistance. 1R eco m m en d atio n s for C h a n g e in the F e d e ra l D e p o s it In su ran ce System , Working Group of the Cabinet Council on Economic Affairs (January 1985), and D e p o s it In s u ra n c e in a C hanging Environm ent, Federal Deposit Insurance Corporation (April 15, 1983). 30 FRBNY Quarterly Review/Autumn 1985 would not, however, extend FDIC insurance coverage to foreign branch deposits. For this article, foreign branch deposits are defined to be both the deposits of foreign and U.S. residents booked at U.S. banks’ offices located overseas and foreigners’ deposits in IBFs and Edge Acts located in the United States. Deposits by foreigners in domestic offices of U.S. banks are already covered under the FDIC insurance system. This article considers a general version of the pro posals. Banks would pay a gross premium rate of onetwelfth of 1 percent on deposits at their foreign branches, the same rate as on deposits at their domestic offices, but would receive no FDIC insurance coverage on these deposits.2 Proponents of imposing FDIC premiums on foreign branch deposits identify two major benefits from the proposed change: a fairer division of the FDIC premium burden and an improved competitive position for small banks relative to large ones. This article will analyze the proposal only in light of these two goals. Equity and competitiveness are desirable characteristics of an effective deposit insurance system, but not its overriding goals. The primary purpose of deposit insurance is to provide a safety net for depositors in the event of a bank failure and thereby to protect the integrity of the 2Banks pay the gross premium rate on their deposits, but the FDIC has always rebated a portion of it at the end of the fiscal year. The gross premium rate less the portion rebated is the net, or effective, premium. banking system. Equity and competitiveness are also not the only goals that have been put forward in the broader discussion of deposit insurance reform. The two goals represent separate issues, which can and should be analyzed separately, as they are in this article. Analysis may suggest accepting one goal but not the other. Considering the goals separately is mean ingful because a deposit insurance scheme can be designed to accomplish both goals, or one goal without the other.3 The first goal, a fairer division of the premium burden, is a matter of equity. The relevant issue is the relationship between the burden borne by an individual bank and the benefits accruing to the bank and its depositors.4 The second aim, improved competitive position for small banks, focuses on the marginal cost of deposit insurance, the premium rate on those liabilities that banks use to adjust their funding on a short-run basis. Here the analysis concentrates on the limited issue of whether large banks face such significantly lower mar ginal deposit insurance costs under the present premium arrangements that they have a competitive advantage over smaller banks in pricing loans. This is not the only bank competitiveness issue raised by deposit insurance. Another, perhaps more important issue relates to depositor perceptions of how deposit insurance coverage applies in practice. Small bank representatives generally maintain that they are at a competitive funding disadvantage because the public views insurance of large bank deposits as more exten sive. The cost consequences of perceptions of deposit insurance coverage are different from the cost conse quences of the deposit base for insurance premiums and are not examined here. A fairer distribution of premiums The first goal of the proposed extension of the premium base is to produce a fairer distribution of the premium burden. And the proposal does substantially redistribute the burden toward large banks. But the proposal’s equity *For example, it could be achieved through a combination of lump sum and marginal insurance premiums. 4This article focuses on one aspect of the fairness of the distribution of premium charges— the relationship of the premium base to insured deposits. There are other aspects of fairness that the proposal does not address and which therefore are not discussed here. Among them is the extent to which differing riskiness of individual banks should be incorporated into the premium structure. A second issue is the extent to which deposit insurance is equally valued by the depositors at small and large banks. Depositors can evaluate the creditworthiness of large depository institutions better than smaller ones because more financial analysis and credit evaluation is available for large banks. For small banks, deposit insurance can substitute for this kind of information. depends on how one views the insurance coverage— this is a matter open to considerable debate. Differing views involve distinctions on two crucial issues: how extensively uninsured deposits are covered and how banks of different types are treated in the event of a failure. The distinction concerning coverage can be described in terms of limited de jure versus more comprehensive de facto insurance coverage. De jure insurance cov erage may be used to denote the insurance explicitly provided by law, which is limited to $100,000 for each depositor.5 De facto insurance coverage, in this dis cussion, refers to the protection uninsured depositors perceive they have, since they may actually suffer no losses when the FDIC merges or sells, rather than liq uidates, troubled institutions. The need to economize and conserve FDIC resources requires minimizing the cost of handling troubled institutions. In the vast majority of cases this has resulted in purchase and assumption arrangements that have maintained the value of all deposits. Even in circumstances where a merger or sale of assets cannot be arranged, other considerations, such as fears of systemic risk and the desire to avoid inter ruptions in depositor service may lead the FDIC to provide more than the legally required deposit protection. A second distinction involves perceptions of how the FDIC treats banks of different types, particularly in the event of a failure. If all banks receive the same treat ment, the system may be termed unified. But if banks fall into two groups according to their size, for example, with uninsured liabilities treated differently if they fail, the system should be described as two-tiered or dual. To highlight the role of these distinctions in evaluating the proposal’s equity, this article examines two very stylized versions of the deposit insurance system. Actual FDIC practice lies somewhere between them. It is important to remember that far more often than not, the practice here and abroad is to merge or sell failing institutions rather than to liquidate them. Thus, unin sured depositors have generally not suffered losses in bank failures. Moreover, the decision to merge or to liquidate is made on a case-by-case basis according to the specific circumstances of the troubled bank, and not just on the basis of a bank’s size, as these highly styl ized versions of coverage might suggest. Therefore, some uncertainty about the extent of de facto coverage exists for all banks, regardless of their size. The caseby-case approach means that depositors probably would technically, coverage is limited to the first $100,000, aggregated over all accounts for each right and capacity of the depositor. This means that an individual can set up separate rights and capacities through joint accounts or trusteeships in addition to his or her individual right and capacity. For corporations, the ability to establish additional rights and capacities through joint tenancy is a matter of controversy. FRBNY Quarterly Review/Autumn 1985 31 not perceive the level of de facto coverage based solely on the observed frequency of mergers or sales in resolving bank failures. The two very stylized views of the insurance system which emerge from these distinctions are: • Deposit insurance coverage as a unified system. Depositors at all banks receive the same de jure protection of insured deposits and no coverage of uninsured liabilities. A variant of this first view perceives a unified system in which as a general practice uninsured depositors at all banks, regard less of size, receive the same de facto coverage of legally uninsured liabilities. • FDIC insurance coverage as a dual system. Legally uninsured as well as insured liabilities are de factocovered at larger banks, but as a general practice only insured deposits are protected at smaller institutions. Since the dividing line between large banks and small banks is unclear, large depositors have an incentive to evaluate carefully the credit worthiness of banks holding their deposits. As the next sections explain, each of these stylized views of FDIC coverage leads to a different assessment of the proposed extension of the FDIC premium base. Under the unified system view, the proposal appears to increase inequity when coverage is only de jure, but as the extent of de facto coverage increases, this effect diminishes. Under the dual system view, the effect of the proposal would be ambiguous. Discrepancy between cost and benefit under the current premium system FDIC insurance protects the first $100,000 of each domestic deposit account at premium-paying banks. In return, banks pay a uniform premium rate of one-twelfth of 1 percent on all domestic deposits, including that portion of deposits over the $ 1 0 0 , 0 0 0 ceiling and thus not covered by FDIC insurance. The FDIC describes this as a “ flat-rate” system, because banks pay the same premium on all domestic deposits. But “ flat rate” may be a misnomer since it suggests that banks pay a uniform price for insurance coverage. In fact, they do not. Based on the cost per dollar of domestic deposits, a bank that relies heavily on large (over $100,000) Certificates of Deposit (CDs) for its funding will pay more for its de jure coverage than a bank with mostly retail deposits under $ 1 0 0 , 0 0 0 each. If the deposit insurance system is viewed as unified and de jure, treating all banks equally and insuring each depositor only up to $ 1 0 0 ,0 0 0 , then the average large bank may subsidize the average small 32 FRBNY Quarterly Review/Autumn 1985 Table 1 Share of Large Deposits at Insured Banks By size of bank, as of June 30, 1984 FDIC-insured banks with assets of: Number of banks Uninsured domestic deposit liabilities* Deposits at foreign branches 0 to $300 million ............ $300 million to $1 billion . $1 billion to $5 billion $5 billion to $10 billion ... Over $10 billion .............. 13,670 453 201 34 23 10.7 19.8 27.7 28.9 22.6 0.1 0.4 7.0 14.1 48.3 All FDIC-insured banks .. 14,381 20.1 18.4 'Calculated as total deposits over $100,000 (large deposits) less $100,000 times the number of large deposits. Source: Call Reports (June 1984). bank (assuming that all banks are equally risky), because proportionally more uninsured liabilities are held at large banks (Table 1 ) . 6 Subsidization may also occur among banks of similar size, since the reliance on uninsured d e p o sits am ong banks v a rie s . For example, some small banks have substantial uninsured deposit liabilities. What if the system is viewed as unified but offering partial de facto coverage for legally uninsured liabilities? According to the FDIC , 7 uninsured depositors assume that they have at least partial de facto deposit protection because the FDIC tends to arrange the merger or pur chase of a troubled or closed bank, rather than its (liquidation. If so, then charging insurance premiums on the legally uninsured portion of deposits can be appropriate, but the premium rate should reflect the extent of de facto coverage, generally less than for fully insured deposits. Under the current premium arrangements, if there is the same partial de facto coverage for all banks, the extent of subsidization of some banks by others becom es unclear. Banks w ith s u b s ta n tia l domestic and few foreign uninsured liabilities still pay more for their coverage than banks with mostly insured deposits, since the premiums do not reflect the different levels of coverage of insured and uninsured deposits, but the disparities are smaller than those under a unified •June 1984 rather than March 1985 data are used because data on insured and uninsured liabilities are collected only once a year on the Call Reports. Uninsured liabilities are measured as the excess of each deposit over $100,000, a somewhat inaccurate measure (see footnote 5 for further reference). 7Deposit Insurance, op. cit. system with de jure coverage only. The situation is less clear for banks with substantial foreign as well as domestic uninsured liabilities. The premiums on the domestic uninsured liabilities may be high relative to the partial coverage they receive, but banks pay no pre miums on the foreign branch liabilities. Thus, whether these banks pay too much or too little for their coverage depends on the level of de facto coverage and the distribution of deposits between foreign and domestic uninsured liabilities. Adopting the dual system view alters the evaluation dramatically. Some observers have suggested that de facto insurance coverage of uninsured deposits at large banks, but only large banks, is widely perceived to be 100 percent. The view is an extreme characterization, but for some it seems to be reinforced by the manner in which the problems of Continental Illinois were han dled last year.8 Perception is inherently hard to ascertain, however. Reasoning very generally that the disruption and drain on the FDIC’s resources in the event of a large bank failure could be too great, depositors may assume that the FDIC would never liquidate in such a case, but would arrange for a purchase or merger into another institution. Large depositors would generally suffer no losses in such a merger.9 Under this view, large depositors in large banks may appear to face less risk than large depositors in small and medium-sized banks. But experience shows that at the first sign of trouble, large depositors may quickly shift deposits to another institution. Such behavior is potentially inconsistent with a perception of full de facto coverage. Under the dual system view, the largest banks pay too little for their insurance, because they do not pay pre miums on their foreign branch deposits which are cov ered de facto. Meanwhile, smaller banks with substantial uninsured domestic deposits pay too much. How equi table the system is to small banks with mostly insured deposits under such a system is unclear; their premiums per dollar of insured deposits could be higher or lower depending on the distribution of uninsured deposits in the dual system’s two tiers. Of course, this analysis ignores any differences in risk among different classes of banks.10 ■The sharp rise in rates paid on Continental Illinois’ and other banks’ CDs during the late spring and early summer of 1984, however, indicates that this perception was not universally held. •A recent proposal by the FDIC to introduce a modified payout (only partial reimbursement) to uninsured creditors could affect these perceptions. ,0But note that the risk-related premium system advocated by the FDIC and the Treasury studies already cited would not correct the discrepancy between the premium base and the amount of coverage. In summary, then, if one analyzes the current premium arrangements according to the stylized unified system view with de jure coverage of legally uninsured liabili ties, banks with sizable uninsured domestic liabilities appear to pay more for their insurance coverage than banks with mostly insured liabilities, assuming they are of equal risk. If all banks have some de facto coverage, banks with uninsured domestic liabilities and no foreign liabilities still appear to pay more for their insurance coverage. Banks with substantial foreign liabilities, however, may pay more or less relative to other banks depending on the extent of the de facto coverage and the distribution between uninsured domestic and foreign deposits. If one accepts the stylized dual system view, small and medium-sized banks with substantial domestic uninsured deposits appear to pay more for their cov erage than large banks.11 Redistribution of premiums under the proposal The proposed extension of the premium base would redistribute premiums substantially (Table 2). Based on March 31, 1985 Call Reports data for 14,379 FDICinsured banks, the major burden of expanding the pre mium base would fall on the 24 banks with assets of $10 billion or more; their combined increase in pre miums would amount to $239 million per year. Another 137 banks with assets between $1 billion and $10 billion would pay $35 million in additional premiums. Among smaller banks, 53 have foreign branch deposits and these banks together would pay $1 million more. The result would be a rise of $276 million in total FDIC premiums, an increase of 21 percent. The proposal as a repricing of FDIC insurance Bringing the deposits of foreign branches into the FDIC premium base can be viewed as a way to reprice the insurance. Comparing the proportion of selected large liabilities before and after foreign branch deposits are included shows how the repricing would work (Table 3). Under the current premium arrangements, the largest banks pay relatively more for their de jure insurance coverage. The de jure protection declines as the share of uninsured domestic deposit liabilities increases—and that share is much higher for large banks than for small banks (Table 3, column 1). Adding the foreign deposits to both the uninsured liabilities and the base produces an even steeper rise in the share. Now, the share rises from 11There are more sophisticated ways to measure the degree of subsidization, including incorporating a measure of the institution’s riskiness. See, for example, Alan J. Marcus and Israel Shaked, “The Valuation of FDIC Deposit Insurance Using Option-Pricing Estimates", J ournal o f M oney, C redit, a n d B anking, Volume 16, No. 4, Part 1 (November 1984), pages 446-460. But as the sophistication of the methodology grows, the possible objections multiply and uncertainty about the validity of the result increases. FRBNY Quarterly Review/Autumn 1985 33 11 percent for the smallest banks to 44 percent for the largest. Under the proposed arrangements, it would range from 11 percent all the way up to 71 percent. It is not just large banks that currently face this kind of gap between the premium base and insured deposits. At 300 banks, the share of uninsured domestic deposit liabilities in all domestic deposits exceeds 40 percent, the average share of these accounts at large banks. Of the 300 banks, more than half have assets of less than $300 million, about 1 percent of all banks in that size class. Under a unified deposit insurance system with the same partial de facto coverage of uninsured liabilities for all banks, to include foreign branch deposits would still leave a gap between the deposit base and insur ance coverage. The size of the disparity would depend on how much partial coverage uninsured liabilities received; it would only disappear when de facto insur ance coverage reached 100 percent. All told, under the stylized unified system view, the proposal would make banks with large deposits pay more for their coverage relative to smaller banks than they do now. However, if one sees the insurance system as dual, the repricing creates different effects. The size of foreign Table 2 FDIC Premiums Under the Proposed Extension of the Premium Base Computed as of March 31, 1985 Millions of dollars Group of banks All insured banks Banks with assets Banks with assets Banks with assets ............................................ of less than $1 billion........ of $1 billion to $10 billion .. of $10 billion or m ore ........ Number in group Number with foreign deposits Domestic deposits Foreign deposits* Current premiumf Proposed premium} 14,379 14,106 249 24 214 53 137 24 1,605,560 789,898 490,838 324,824 330,702 1,422 42,044 287,237 1,338.0 658.2 409.0 270.7 1,613.6 659.4 444.1 510.1 Difference 275.6 1.2 35.1 239.4 "Deposits at foreign branches, Edge Acts, and International Banking Facilities. fOne-twelfth of 1 percent of domestic deposits. ^One-twelfth of 1 percent of total deposits. Source: Call Reports (March 1985). Table 3 Proportion of Selected Large Deposits in the Premium Base As of June 30, 1984 Using domestic deposits as the __________________________ premium base (1) FDIC insured banks with assets of: Number of banks Uninsured domestic deposit liabilities* (2) Foreign branch deposits Using all deposits as the premium base (3) Uninsured domestic deposit liabilities plus foreign branch deposits 0 to $300 million .......................................................... $300 million to $1 billion ............................................. $1 billion to $5 billion .................................................. $5 billion to $10 billio n ................................................ Over $10 b illion........................................................... 13,670 453 201 34 23 10.7 19.9 29.8 33.7 43.7 0.1 0.4 7.6 16.3 93.5 10.8 20.2 34.7 43.0 70.9 All FDIC insured b a n k s f............................................. 14,381 24.6 22.5 38.5 ’ Calculated as all deposits over $100,000 (large deposits) less $100,000 times the number of large deposits. fSince the large banks dominate the average, especially after the inclusion of foreign deposits, a comparison of the large bank proportion to the average is not very meaningful. Source: Call Reports (June 1984). 34 FRBNY Quarterly Review/Autumn 1985 deposits relative to the base provides an indicator of the amount of excess de facto coverage large banks now receive. The foreign branch deposits of banks with assets over $1 billion are substantial, relative to the present premium base, and jump sharply with bank size (Table 3, column 2); for the top 23 banks, foreign branch deposits nearly equal all domestic deposits. Under the dual system view, these very large banks would wind up paying less for their actual coverage than smaller banks because the FDIC to some extent protects foreign branch deposits of large banks.12 Including foreign branch deposits redistributes, but does not eliminate, the discrepancy between the base on which premiums are charged and the deposits cov ered by insurance, under the dual system view. The revised premium base narrows the gap for any banks viewed as being in the first tier which has some de facto coverage, eliminating it only if the de facto cov erage is 100 percent. But for banks considered to be in the second tier, adding foreign branch deposits has the same effect as the unified system view implies: it creates a sharp rise in large banks’ share of uninsured liabilities in their premium base. For the 34 banks with assets between $5 billion and $10 billion, the share increases from 34 percent to 43 percent, while for the 23 largest banks, it jumps from 44 percent to 71 per cent. Among banks with assets under $1 billion, foreign branch deposits are so small that including them makes little difference. To sum up, the proposed extension of the premium base cannot produce an unambiguously fairer distri bution of the FDIC premium burden, no matter which of the two views of the deposit insurance system one accepts. These stylized views should help to highlight how differentiation in the treatment of banks and in the extent of de facto coverage influence the fairness of the proposed redistribution. Under the unified system view, the proposal only exacerbates the disparity between the premiums paid and the deposits insured, unless de facto coverage is thought to be total. Even under the dual system view, the change does not fully align pre miums with the perceived differences in coverage between the dual system’s two tiers of banks because the first tier (with de facto insurance) is not distin guished from the group of banks with large uninsured and foreign branch deposits. The proposed redistribution will not be fair to some members of the latter group. An arrangement that imposes premiums by deposit type, rather than bank type, charges some banks for coverage they will not get under the dual system view. Indeed, a full evaluation of the equity of the proposal under the 12Some of these uninsured deposits are liabilities to other U.S. banks, as they are in the domestic market. dual system view would require an explicit definition of the first and the second tiers. The inherently arbitrary nature of such a distinction underscores the extreme character of the dual system view. Improving the competitiveness of small banks The second goal of a proposed extension of the FDIC premium base is to improve the competitive position of small domestic banks relative to large ones. To accomplish this, the proposal tries to equalize the marginal cost of deposit insurance across all deposit types for all U.S. banks.13 The change would tend to raise the marginal cost of funding for large banks relative to small ones. Applying an FDIC premium to deposits at foreign branches would equalize the marginal insurance cost (but not neces sarily the total marginal cost) on international and domestic deposits. Funding costs for U.S. banks in the international markets would increase, because the highly competitive nature of those markets would prevent U.S. banks from passing on much of the increased cost to their deposit customers. If the new relative funding costs then get incorporated into loan pricing, the cost of loans at large banks with access to the Euromarket would rise relative to that of small banks with a purely domestic base. The change would in theory tend to shift market share of total loans and deposits held by U.S. banks toward small banks and away from large banks. The size of the impact would depend on how much small funding cost differences determine market struc ture in the banking industry. Research on this question suggests that other factors—such as regulation, econ omies of scale in providing certain services, and advantages gained by specializing in particular ser vices—play an important role in the structure of com petition between large and small banks.14 This literature emphasizes that local banking markets are small; as a consequence, regulatory control of entry and branching is very important. Further, cost savings may arise from the joint production of several banking services. By contrast, the funding cost advantage of access to the Euromarkets has received little or no weight. Therefore, 13Differences in marginal insurance premiums are only a part of the difference in marginal funding costs across banks, so the proposal would not equalize marginal funding costs for all banks. 14See, for example, George J. Benston, Gerald A. Hanweck, and David B. Humphrey, “Scale Economies in Banking: A Restructuring and Reassessment”, Jo u rn a l of M oney, C re d it, a n d B anking, Volume 14, No. 4, Part 1 (November 1982), pages 435-456; Thomas Gilligan, Michael Smirlock, and William Marshall, “Scale and Scope Economies in the Multi-Product Banking Firm”, J o u rn a l o f M o n e ta ry Econom ics, Volume 13, No. 3 (May 1984), pages 393-405; and Sherill Shaffer, "Competition, Economies of Scale, and Diversity of Firm Sizes”, A p p lie d E conom ics, forthcoming. A number of studies are summarized in R. Alton Gilbert, “Bank Market Structure and Competition”, Jo u rn a l o f M oney, C redit, a n d B anking, Volume 14, No. 4, Part 2 (November 1984), pages 617-645. FRBNY Quarterly Review/Autumn 1985 35 small changes in relative funding costs alone are unlikely to have any great effect. Altogether, the degree of competition among banks of similar size is quite possibly greater than that among banks of different size. Some observers have argued that perceived differ ences in bank safety are a major factor affecting com petition. Since the proposal does not include formal extension of FDIC coverage to deposits at foreign branches of U.S. banks, implementing it should not alter these perceptions. However, the analysis of the impact of FDIC premiums on market terms and market shares would be different if foreigners and U.S. residents viewed deposits in for eign branches of U.S. banks as effectively having more insurance protection than before, notwithstanding the lack of formal (de jure) coverage. Such reassurance could be quite valuable. The normal tiering in the Euromarket suggests that safety may be worth more than 8 basis points, the increase in cost from imposing FDIC premiums on foreign branch deposits. Extending FDIC insurance premiums to foreign deposits of U.S. banks may not give such a clear signal to market participants, however. Extending the base appears consistent with the dual insurance system view by implying that some de facto coverage for large deposits at international banks already exists. But important features of that system remain unspecified, particularly the boundary between banks with some protection of uninsured liabilities and those without it. Foreign branch depositors would be left uncertain about just how much of their deposits would be covered in a bank failure—as is now the case. U.S. competitiveness in domestic markets and abroad The proposed change in premium structure could alter the competitive structure of banking in the United States and abroad. To begin with, applying an FDIC premium to foreign branch deposits would raise the cost of external funds. Under the assumption that the FDIC would rebate nothing from the gross premium, the effective rate of premium would be one-twelfth of 1 percent or 8.3 basis points.15 For banks subject to the 3 percent reserve requirement on Eurocurrency liabili ties, the effective cost of external funds would rise 8.6 basis points.16 These are small changes compared with the daily volatility of Eurodollar rates, for example, which 15The FDIC rebate has declined in recent years; it rebated only 13.5 percent of the premium to the banks in 1983 compared with as much as 60 percent earlier. '•For banks subject to reserve requirements on Eurocurrency liabilities, the effective cost of external funds is: iE$ + FDIC iE$ + .083 1 - RRe$ ~ 1 - .03 where iE$ is the relevant Eurodeposit offer rate (e .g ., three months), 36 FRBNY Quarterly Review/Autumn 1985 in 1984 averaged 140 basis points when measured by the standard deviation. However, these small changes are large relative to current Euromarket margins. Fur ther, the change would create a permanently higher average cost of external funds and their effects would tend to persist. Higher external funding costs could place modest upward pressure on domestic funding costs and lending rates. Applying an FDIC premium to foreign branch deposits would reduce the competitiveness of U.S. banks with foreign deposits relative to non-U.S. banks operating in the Euromarkets. The increased cost of external funds, relative to domestic funds, would lead large banks to adjust their marginal funding from foreign to domestic markets, especially since they might not be able to shrink assets rapidly enough in response to declines in liabilities. Impact on the market shares of U.S. banks at home and abroad How much market shares in the Eurocurrency and domestic lending markets change would depend on how market terms responded to a shift in U.S. bank funding costs. While the cost differences would be small, they would be large relative to current Euromarket margins, and since the volumes are large, the size of the impact cannot be determined precisely. However, since the Eurocurrency market is highly competitive, flows might well be significantly redirected. In the domestic market, higher marginal funding costs could lead large banks to price loans higher, at least on the parts of their loan portfolio with thin profit margins. On loans with higher profit margins, the banks might instead absorb all the funding cost increase. Smaller domestically-funded banks, with lower marginal funding costs, could build up profits or quote slightly lower loan costs. That would push market share toward small banks. In the Euromarkets where profit margins are already thin, more expensive funds would probably impel U.S. banks to quote less favorable terms. Since U.S. banks form a large segment of the market, foreign banks would find themselves attracting depositors and bor rowers in the Euromarkets away from U.S. banks, thus increasing their market share. Of course, foreign banks would only be willing to expand their Eurocurrency balance sheets at current interest rates if they faced no legal or internal balance sheet constraints.17 In the short run, such constraints Footnote 16, co n tin u ed FDIC is the premium rate, and RREt is the reserve requirement on Eurocurrency deposits. 17Another possibility is that foreign banks not now active in the Euromarket would enter. This seems less likely now than it would (p .3 8 ) How Interest Elasticity Affects Deposit Losses rates at non-U.S. Euromarket banks fall, their decline would blunt the impact of the FDIC premium. Only at fairly high elasticities would the deposit losses and FDIC revenue reductions become substantial. The deposit losses range from one-tenth of 1 percent if rates were high and the elasticity low, to about 17 percent at an interest rate of 5 percent and an elasticity of ten. The maximum loss of revenue to the FDIC on the table is $46 million, still less than one-third of the amount rebated for 1983. Larger declines are possible if the interest elasticity of foreign branch deposits is higher.f The relatively small share of branch deposits in total deposits limits the maximum possible revenue loss through this channel to about 17 percent of revenues, the share of foreign deposits in total deposits. Assuming no FDIC rebate, not just the foreign but the domestic deposit base could also erode. FDIC revenue shortfalls would eventually require higher FDIC pre miums, which would lower domestic deposit rates in the United States. The interest sensitivity of domestic deposits in aggregate is likely to be less than that of foreign deposits, since domestic deposits include small transactions accounts and time deposits with low interest elasticity along with highly interest-sensitive ones. But since the base of domestic deposits is much larger, even modest declines in deposit rates following an FDIC pre mium increase could produce very substantial revenue losses. The extent of deposit losses under the proposed exten sion of the premium base would depend on the interest elasticity of deposits and the level of interest rates. Estimating these losses requires knowledge of depositor interest sensitivity, and the overall level of interest rates. The elasticities are difficult to measure, since the small differences in rates to which banks and depositors respond are not observable without continuous data collection on interest rates over the day. One can only infer that the elasticity is quite high. Sample computations provide some idea of the mag nitude of deposit losses and revenue shortfalls under different assumed depositor interest elasticities and levels of interest rates in the Euromarket (table). Interest elasticities can range from zero (interest insensitivity) to infinity.* The elasticities here reflect a range of low to high, but it is quite likely that foreign branch deposits are even more interest-sensitive than implied by the interest elasticity of ten. The range of the interest rate is rep resentative of Eurodollar rates over the last ten years. The computations here assume that imposing an FDIC premium on deposits at foreign branches of U.S. banks would have no effect on U.S. domestic rates or on deposit rates at non-U.S. Euromarket banks. If deposit *The interest elasticity gives the percentage decline (increase) in deposits for a 1 percent decline (increase) in interest rates. Interest sensitivity increases as the elasticity rises in value. As it approaches infinity, small changes induce depositors to withdraw all their deposits and invest them in an alternative instrument. fThe revenue losses will increase proportionally with the elasticity (e.g., an elasticity of 20 will produce double the revenue decline of an elasticity of ten). FDIC Revenue Reductions Under Alternative Interest Rate and Interest Elasticity Assum ptions In millions of dollars Assumed interestelasticity Domestic deposits Foreign deposits Premiums under proposed premium base extension Reduction of premium from base case * 1,605,560 330,702 1613.6 * Interest rates of 5 percent ................ ............ 0.2 1.0 10.0 1,605,560 1,605,560 1,605,560 329,600 325,190 275,585 1612.7 1609.0 1567.7 -0 .9 -4 .6 -45 .9 Interest rates of 10 p ercent.............. ............ 0.2 1.0 10.0 1,605,560 1,605,560 1,605,560 330,151 327,946 303,144 1613.1 1611.3 1590.6 -0 .5 -2 .3 -2 3 .0 Interest rates of 15 p ercent.............. ............ 0.2 1.0 10.0 1,605,560 1,605,560 1,605,560 330,335 328,865 312,330 1613.3 1612.1 1598.3 -0 .3 -1 .5 -15.3 Current (March 1985)....................... 'Not applicable. FRBNY Quarterly Review/Autumn 1985 could leave foreign banks little choice but to adjust to some of the impact of higher U.S. bank funding costs in larger spreads and higher profit margins. But even tually, accumulated capital from those higher profits would ease the balance sheet constraint and allow for eign banks to pursue a larger market share. Similarly, foreign banks would only expand their balance sheets at current rates if the marginal costs of loan production do not rise too sharply. Credit evaluation and loan servicing costs may be higher for loans to new bor rowers than for their normal loan portfolio. This would lead foreign banks to compensate by increasing their spreads charged over LIBOR, possibly eliminating their competitive advantage. But experience and economies of scale may allow spreads to narrow in the longer run. In summary, the extent to which U.S. banks would lose market share and bid-offer spreads would widen depends mainly on two things: the willingness of foreign banks to increase their Eurocurrency balance sheets, and the interest sensitivity of depositors, borrowers, and lenders. Euromarket participants could, of course, shift their activities to other markets as well as to other agents in the Euromarket. In general, the more willing foreign banks are to increase balance sheets and the more sensitive market participants are to interest rates, the greater U.S. losses in market share would be and the smaller changes would be from current market terms. These effects would be mitigated if depositors perceived greater coverage for their funds in foreign branches or reinforced if depositors became more uncertain of the extent of coverage. Consequences of a falling U.S. market share Any reduction of the U.S. banks’ share of the domestic or Eurodeposit market would tend to shrink the deposit base on which premiums would be charged, assuming no growth in deposits. The magnitude of the decline is difficult to judge, but the possibility that it could be siz able cannot be ignored.18 Moreover, the deposit Footnote 17, c o n tin u ed have been in the 1970s, when participation in the Euromarket was increasing rapidly. ’•The decline in the deposit base does not necessarily have to reflect a shrinking U.S. share of world bank liabilities. Financial innovation, in the form of new non-deposit liabilities, could follow a rise in insurance premiums on Eurodeposits. The proposed premium could also further encourage the growth of off-balance sheet transactions by banks. shrinkage does not imply that the FDIC would imme diately need less funds. In the long run the FDIC’s exposure should decline along with the deposit base (assuming the deposit base does not fund riskier assets). In the short run, however, its exposure reflects past experience. Thus, the financing cushion the proposed change would provide to the FDIC may be smaller than expected. With a substantial erosion of the deposit base, revenues from the foreign branch deposits might not be as high as projected (box). Currently, the FDIC rebates the excess premium paid, and this allows some margin for the inevitable error in gauging its needs and revenues. That margin has been disappearing, though, and the rebate has shrunk. Conclusion Extending the premium base for FDIC insurance to deposits at the foreign branches of U.S. banks would raise FDIC revenues by 21 percent and substantially redistribute deposit premiums from small and medium sized banks to large ones. Whether this redistribution is appropriate depends largely on how one views the extent of de facto coverage and the unity of treatment of banks of differing characteristics, including size. At one extreme, if one accepts the dual insurance system view that large banks regularly receive more de facto insurance protection than small banks, then large banks would in fact be paying more for the effectively higher coverage they receive. At another extreme, if one views the system as unified, the proposal would raise the insurance cost per dollar of insured deposits to all banks with deposits at foreign branches. This is true whether all banks tend to receive the same partial de facto coverage of uninsured deposits or none at all. But the proposed change would not eliminate the discrepancy between the premium base and the amount of insurance coverage for all groups of banks, under either stylized view of the system. For many medium-sized and fairly large banks with foreign deposits, the proposal may widen the gap substantially. The competitive implications of the proposal also raise questions. Equalizing the marginal insurance cost of funds between the Euromarkets and the domestic money markets for U.S. banks would necessarily raise the funding costs of U.S. banks relative to those of other banks in the Euromarkets. Christine M. Cumming 38 FRBNY Quarterly Review/Autumn 1985 ARMs: Their Financing Rate and Impact on Housing When widespread use of adjustable rate mortgages (ARMs) was permitted in April 1981, some analysts expected housing demand to become stronger and less sensitive to interest rate fluctuations as prospective homebuyers turned to this new way of financing homes. Because housing is one of the most interest-sensitive sectors of the economy, this effect could influence the dynamics of the business cycle and the countercyclical effectiveness of monetary policy. Recent evidence suggests, however, that ARMs have not had a large impact on housing demand. This seems paradoxical because ARMs have captured a large share of new mortgages, particularly between mid-1983 and mid-1984. We offer a twofold explanation for this paradox. First, we show that ARMs have in effect generally not been priced much lower than fixed-rate mortgages (FRMs). Second, we examine some characteristics of ARMs that may explain their popularity over FRMs as a mode of finance, even though these features have not signifi cantly increased the incentives to purchase a home. Econometric evidence on housing demand In several recent studies, analysts have found that adding variables representing ARMs contributes little, if any, tracking power to traditionally specified models of housing demand.1 For example, the equation specified The authors would like to thank the following individuals for their comments and suggestions: M. A. Akhtar and A. Steven Englander (Federal Reserve Bank of New York); James L. Freund and John L. Goodman (Board of Governors of the Federal Reserve System); Michael J. Lea (Federal Home Loan Mortgage Corporation); and Randall J. Pozdena (Federal Reserve Bank of San Francisco). 'S ee Howard Esaki and Judy Wachtenheim, "Explaining the Recent Level of Single-Family Housing Starts”, this Q uarterly R ev ie w (Winter by Esaki and Wachtenheim, which has no ARMs vari able, has a post-sample (1982-1 to 1984-IV) mean absolute error of 77,000 units, about 8 percent of single family housing starts (Table 1). And it shows no con sistent tendency to underpredict, a tendency that would indicate a failure to capture the positive influence of ARMs in the housing market; instead the equation mostly overpredicts. The tracking performances of the Esaki/Wachtenheim and other recent models suggest that at most ARMs have had a minor impact on housing demand. The econometric approach, however, may be of lim ited value for analyzing the impact of this financial innovation since there is not a long series of consistent data on ARMs. Thus, we obtain independent confir mation of these analysts’ results by evaluating the long term expected financing rate of a mortgage, i.e., the average rate an owner expects to pay over the period of home ownership. If the long-term financing rate of ARMs has been significantly below the FRM rate, then housing demand should have been boosted substantially. The financing rate of mortgages The financing cost underlying the demand for housing is the interest an individual expects to pay over the Footnote 1, con tin u ed 1984-85), pages 31-38; James L. Freund, "A Small Econometric Model for Predicting Residential Construction Activity: Some Preliminary Results", Board of Governors of the Federal Reserve System, paper presented at the 1984 meeting of the American Real Estate and Urban Economics Association; and Michael J. Stutzer and William Roberds, “Adjustable Rate Mortgages: Increasing Efficiency More Than Housing Activity", Federal Reserve Bank of Minneapolis Q uarterly R eview (Summer 1985), pages 10-20. FRBNY Quarterly Review/Autumn 1985 39 period of ow nership. In the case of an FRM, the expected cost, excluding the initial points that exist also for ARMs, never exceeds the amount determined by the contract rate and may be lower if market rates fall enough to make refinancing advantageous. With an ARM, the expected cost is contingent upon future short term rates. Thus, the first-period rate discount of an ARM is only one element of the total financing cost, and has to be considered along with the likely course of future rates and the expected holding period in judging the costs of an ARM. Nonetheless, some analysts believe that this relatively low first period rate of ARMs boosts housing demand, regardless of the expected course of interest rates, by permitting more people to qualify for a mortgage. Many more individuals will meet a stipulated maximum limit on the share of income earmarked for mortgage pay ments if the first-year ARM rate instead of the FRM rate is used to calculate the carrying costs for a prospective borrower . 2 From the lenders’ perspective, relaxing screening procedures may have been one way to encourage a faster reshuffling of their portfolios from FRMs to ARMs; the lower interest rate risk of ARMs to lenders may more than compensate for the higher credit risk. Moreover, some market observers say that lenders may have eased qualification criteria in the belief that the default risk is carried by mortgage insurers and repurchasers. Some of these insurers and repurchasers, however, have recently responded by encouraging or requiring lenders to tighten their qualification criteria for ARMs.3 Independently, borrowers may be “ self-policing” by avoiding a commitment that might have a high risk of default .4 On balance, the extent of the effects of the ARM qualification criteria on housing demand are not clear. The low initial ARM rate also might raise housing demand through its effect on the pattern of mortgage payments over time. When the market yield curve is upward sloping, the early years’ payments with an ARM 2See John L. Goodman, Jr., "Adjustable Rate Home Mortgages and the Demand for Mortgage Credit", Board of Governors of the Federal Reserve System, presented at the 1984 meeting of the American Real Estate and Urban Economics Association. He shows that the use of a 10 percent first-year ARM rate allows 38 percent of households to qualify for a mortgage, while a 13.5 percent FRM rate allows only 25 percent to qualify. Both are representative rates for the period July 1983-May 1984. Esaki and Wachtenheim, op. cit., though, do not find that a variable representing such an ARMrelated reduction of mortgage carrying costs helps their econometric model predict single-family housing starts. 3See Dennis Jacobe, "Mortgage Insurers Mix ARMs and GPMs to Justify Rates", Savings Institutions (October 1984), pages 41-45. 4See John L. Goodman, Jr., op. cit., for evidence supporting this view. 40 FRBNY Quarterly Review/Autumn 1985 Table 1 Tracking Performance of the Esaki/Wachtenheim Econometric Model of Housing Demand* Predicted less actual; thousands of units at an annual rate Post-sample prediction errors 1982-1 ......................................... 1982-11 ....................................... 1982-111 ...................................... 1982-IV 40 122 59 4 1983-1. 1983-11 1983-111 1983-IV 60 143 140 12 1984-........................................... 1 1984-11 ....................................... 1984-111 1984-IV -9 2 94 105 57 Positive errors indicate overprediction, i.e., predicted level exceeds actual level. ‘ Howard Esaki and Judy A. Wachtenheim, op. cit. The equation tracks single-family housing starts. The sample period is 1959-IV—1981 -tV, and the mean absolute error of the sample period is 55. are less than with an FRM, but payments are likely to be higher in later years. Similar to the advantages of graduated payment mortgages, this tim ing of ARM payments might be desired by some people because they feel that their incomes are also likely to rise in the future. In this case, the carrying burden of a mortgage may be more uniform over time instead of being heavier initially as it is with an FRM. It is not clear, though, whether this feature of ARMs, by itself, would signifi cantly boost housing demand. Esaki/Wachtenheim, for instance, do not find that variables representing the different payment streams of ARMs and FRMs, e.g., the spread between the FRM and initial ARM rates, help their equation track housing in recent years. Moreover, basing a purchase decision solely on this consideration would be risky given the uncertainty of future ARM rates. In any case, the long-term expected financing rate of ARMs is likely to be a key element in a home purchase decision. However, individuals’ expectations of future rates—the main component of this expected financing rate—are not observable. And there is no consensus on how these expectations are formed. Some analysts believe that people base their expectations on the most recent movements of rates. Others believe that individ uals tend to accept the expectations built into the market yield curve, i.e., the relationship between long- and short-term rates.5 For example, when long-term rates exceed short-term rates, people generally expect that short-term rates will increase but on average will be equal to the current long-term rate. This second viewpoint may well describe a prospective homebuyer. Because a house represents a large share of a typical homeowner’s total assets, the consequences of basing a purchase decision on wrong expectations can be quite costly. To reduce this risk, people probably are most comfortable relying on market expectations in making the decision. Our analysis of the long-term financing rate of ARMs, thus, is based on the assumption that the market yield curve essentially represents the average of expected future interest rates held by prospective homebuyers. Whether most borrowers view the long-term financing rate of ARMs as being higher or lower than the FRM rate, therefore, depends on how lenders price ARMs and FRMs relative to the market yield curve. The rel ative pricing of these mortgages, in turn, depends on the net balance, from the lenders’ perspective, of the risks and other characteristics of each type of mortgage. Specifically, ARMs are more attractive than FRMs to lenders because they eliminate or reduce risks related to balance sheet considerations—i.e., the possibility of lower income when the return from mortgage-type assets does not rise as quickly as the cost of funds to a lender— and mortgage prepayment. On the other hand, increased credit risk, a less developed secondary ARM market, and interest rate caps may push up the relative cost of ARMs. One important factor that would cause lenders to lower the financing rate of ARMs is the shift of interest rate risk to the borrower. If the expected financing rate of an ARM, however, is below that of an FRM only because of this shift, then ARMs would not boost housing demand: the risk of greater-than-expected increases in rates still would have to be compensated for by the return from home ownership. Indeed, given the size of investment a home purchase represents, as well as the substantial costs and discomfort of having to default if rates climb much higher than expected, individuals might require a relatively large cut in the ARM rate to compensate them for assuming the interest rate risk. In other words, a significant increase in •For an analysis of “term structure" theory, see Franco Modigliani and Robert J. Shiller, "Inflation, Rational Expectations, and the Term Structure of Interest Rates", E co nom ica (February 1973), pages 12-43. Recent tests indicate some slight variation in the behavior of interest rates from that implied by term structure theory. However, this variation might be explained in terms of a variable risk premium in long-term rates, which would not be inconsistent with our approach to analyzing ARMs. See Robert J. Shiller, John Y. Campbell, and Kermit L. Schoenholtz, "Forward Rates and Future Policy: Interpreting the Term Structure of Interest Rates", B rookings Papers on E co nom ic A ctivity I (1983), pages 173-223. housing demand might result only if lenders price ARMs much below FRMs. On the basis of the analysis which follows, however, we conclude that, at least since the start of 1984, the net effect of the various factors that distinguish an ARM from an FRM has been small. That is, the long-term expected financing rate of an ARM for most people has been about the same as an FRM. To arrive at this result, we first look at the various factors underlying ARM pricing. Balance sheet considerations of lenders By reducing the interest rate exposure of an entire asset portfolio, ARMs may significantly improve the viability of thrift institutions since for tax purposes they are required to hold a large portion of their assets as mortgages.* When financial deregulation, particularly the phasing out of Regulation Q, allowed rates on deposits to vary with market conditions, the large concentration of FRMs in these institutions’ assets made them vulnerable to substantial income losses when interest rates rose.7 ARMs permitted a better match between their return on assets and their cost of funds. This may be an addi tional gain beyond the reduction of interest rate risk inherent in each mortgage, and thus may persuade these lenders to price ARMs attractively. Lenders that are not required to hold mortgages in their portfolios, e.g., commercial banks, credit unions, and insurance companies, presumably were less affected by the introduction of ARMs. If in response to financial deregulation these institutions chose to hold fewer fixed-rate instruments, they had a broader choice of variable-rate assets, e.g., commercial loans, from which to select. The major impact of ARMs on these lenders may have been to maintain their presence in the mortgage market, thus helping to prevent mortgage rates from rising relative to other interest rates. It is not surprising, then, that thrift institutions have been the most active lenders of ARMs. In 1984, for instance, ARMs accounted for about two-thirds of the mortgages originated by thrifts, but less than 40 percent of those issued by commercial banks.8 •See Robert Van Order, “A Simple Model of Variable-Rate Mortgages", H ousing F in an ce R ev ie w (July 1982), pages 299-311. Because of the large losses sustained by many thrift institutions in recent years, some have enough loss carryover that they do not pay any taxes. As a result, these institutions do not feel compelled to hold the required portion of their portfolios as mortgages. Nevertheless, according to Flow of Funds data, mortgages (including ARMs) and U.S. government agency issues (mostly mortgage pass through securities) constituted substantially more than half of thrift institutions' assets during 1984. 7Some thrift institutions have addressed this interest rate risk by hedging in futures markets and engaging in interest rate “swaps". These activities, though, have not been widespread. •See Federal Home Loan Bank Board, N ew s (February 4, 1985). FRBNY Quarterly Review/Autumn 1985 41 Prepayment risk Lenders also may price ARMs more favorably than FRMs because of the reduced risk of borrowers pre paying before maturity. Since individuals are often per mitted to prepay a mortgage at face value without penalty, the expected return from an FRM is uncertain even though its rate is fixed. The FRM rate, therefore, may embody a charge to cover this uncertainty.9 In contrast, ARMs are less likely to be prepaid when market interest rates fall since their rates, assuming there are no binding caps, would decline as well. Moreover, even if an ARM is prepaid, its rate would likely be the same as that on the newly issued ARM that replaces it. Thus, ARM rates are likely to contain no prepayment premium, or at most one that is not as large as that embodied in the FRM rate. Credit risk Other factors, however, may reduce the attractiveness of ARMs to lenders. Both ARMs and FRMs are vulner able to the typical factors behind borrower default, e.g., cuts in income and net worth, but ARMs are also sub ject to rising interest rates, which may raise the prob ability of default. The prospect of higher interest rates in the future does not necessarily mean that defaults on ARMs will increase, particularly if the rise in rates is a result of higher inflation. In this case, most household incomes should expand as well, permitting borrowers to handle the larger carrying costs of ARMs. Indeed, to the extent that lenders use some measure of the long-term expected financing rate of ARMs (which embodies expectations of future rates) to screen borrowers, the default risk may be kept down. Relatively tough quali fication criteria and rate caps also may help reduce this risk. Nevertheless, future interest rates might rise sub stantially more than was expected when the loan was originated and result in an increase in defaults, partic ularly if the increase in rates is not matched by com parable income gains. So far, defaults on ARMs do not appear to be a major problem. Since January 1985, when separate data on ARMs were first reported, the ARM delinquency rate has been below that of FRMs, possibly because interest rates were falling.10 Nonetheless, ARMs may not always •The risk of prepayment is an important consideration in the pricing of a mortgage. See Henry J. Cassidy, "Selection of an Index for Variable Rate Mortgages", J o u rn a l o f R etail B anking (Winter 1982), pages 27-36; Alden L. Toevs and Jeffrey H. Wernick, "Hedging Interest Rate Risk Inclusive of Prepayment and Credit Risks”, Id e n tific a tio n a n d C ontrol o f Risk in the Thrift Industry, Federal Home Loan Bank of San Francisco, Proceedings of the Ninth Annual Conference (Decem ber 1983), pages 97-122. ’"Delinquency rate data were obtained from the U.S. League of Savings Institutions. These data, however, may be biased against 42 FRBNY Quarterly Review/Autumn 1985 have the better record, particularly if interest rates climb steeply. For example, the default rate for ARMs could jump sharply if their rate rises faster than individuals’ incomes, particularly among borrowers with relatively little accumulated equity in their homes.11 Thus, the credit riskiness of ARMs may represent a potential problem. Mortgage liquidity Another factor that could impinge on the advantages of ARMs to lenders is the absence of a large secondary market for these mortgages. As a result, ARMs are much less liquid than FRMs, for which a well-developed secondary market exists. According to market observers, the growth of a secondary market has been slow because ARMs lack uniformity and because investors are concerned that ARMs may carry more credit risk than FRMs. Caps on ARM rates Unlike the other characteristics of ARMs that affect either borrowers or lenders, caps on the periodic change and life-of-loan level of ARM rates affect both borrowers and lenders. For instance, while these caps may prevent the return on ARMs from keeping pace with a lender’s cost of funds, they also reduce the interest rate risk for a borrower. Consequently, even if caps increase ARM rates, borrowers may be willing to pay for these safeguards. The value of caps depends on the course of future interest rates. Thus, an ex ante valuation should be based on the yield curve. When the yield curve is steep, indicating that interest rates are likely to rise sharply in the future, caps should be worth more to a borrower. In addition, caps would be more valuable to the extent that they prevent an initial ARM rate reduced by a firstperiod discount from climbing to the fully indexed level after the first period.12 At the other extreme, when the yield curve is downward-sloping, a cap on the periodic change in an ARM rate may have negative value to borrowers if it prevents an ARM rate from falling as much as market interest rates. In principle, borrowers and lenders can value caps Footnote 10, co n tin u ed FRMs because no adjustment is made for the length of time mortgages are in existence. Since the FRMs in this sample were outstanding for more years than ARMs, they, according to market observers, are more prone to default. 11See Peggy J. Crawford and Charles P. Harper, "The Effect of the AML Index on the Borrower", H ousing Fin a n c e R ev ie w (October 1983), pages 309-320. See also Robert M. Buckley and Kevin E. Villani, “Problems with the Adjustable Rate Mortgage Regulations”, H ousing F inance R ev ie w (July 1983), pages 183-190. 12The initial period pricing of an ARM is the sum of three parts. The first element is an index rate, e .g ., the one-year Treasury (p . 4 4 ) Valuation of Caps To estimate the value of caps, we analyze how ARMs would have behaved with and without caps if they had been available through the 1970s. By determining, ex post, how the financing costs would have differed with varying discounts and caps, we hope to capture the current ex ante expectations for these ARM modifications. Two horizons for expected home ownership are con sidered: three and eight years. The eight-year horizon represents the average duration of a mortgage,* while the three-year horizon is applicable to about one-quarter of homebuyers, those who expect to resell quickly, f In each case, the fully-indexed ARM rate was assumed to equal 2.8 percentage points above the one-year Treasury rate and to adjust every twelfth month. The use of a ’ Frederick E. Balderston, op. cit. fJohn L. Goodman, Jr., op. cit. constant markup and the one-year Treasury rate as a representative index are consistent with recent surveys.* Simulations of hypothetical ARMs, with and without caps, were run starting in 1970 for each month for which there was data, i.e., ending in 1977 with the eight-year horizon and in 1982 with the three-year. From the simulation results we can find the discounted present values of caps in each month.§ First, we cal culate the present value of the mortgage payments, |The first survey was taken in November 1984; see The Primary Mortgage Market, Federal Home Loan Mortgage Corporation (January 1985). The later survey, taken in February 1985, is unpublished. §Our technique is similar to one developed independently by Patrick H. Henderschott and James 0. Shilling, Valuing ARM Rate Caps: Implications of 1970-84 Interest Rate Behavior, unpublished paper, Ohio State University. Effective Values of Caps In percentage points A: 8-Year Horizon Effective value of: Yield curve slope First-period discount Lifetime cap: 5% Annual cap: 2% Both caps Low .................... <0.5 0.0 1.0 2.0 3.0 0.00 0.01 0.06 0.17 -0 .0 2 0.04 0.11 0.19 0.00 0.06 0.16 0.30 Middle ................ 0.5-1.5 0.0 1.0 2.0 3.0 0.13 0.24 0.39 0.61 0.20 0.26 0.34 0.44 0.29 0.41 0.59 0.88 H ig h .................... >1.5 0.0 1.0 20 3.0 0.68 0.93 1.24 1.59 0.43 0.49 0.64 0.89 0.73 1.04 1.36 1.78 Yield curve slope First-period discount Lifetime cap: 5% Annual cap: 2% Both caps Low .................... <0.5 0.0 1.0 2.0 3.0 -0 .0 4 0.00 0.01 0.02 -0 .2 2 -0 .1 3 -0 .0 5 0.02 -0 .2 2 -0 .1 3 -0 .0 5 0.02 Middle ................ 0.5-1.5 0.0 1.0 2.0 3.0 0.00 0.01 0.02 0.06 0.00 0.04 0.12 0.27 0.00 0.04 0.12 0.27 H ig h .................... >1.5 0.0 1.0 2.0 3.0 0.06 0.12 0.21 0.35 0.27 0.42 0.61 0.81 0.27 0.42 0.61 0.81 Group* B: 3-Year Horizon Effective value of: Group* ‘ Each group consists of one-third of the simulation results, ranked by the magnitude of the value of both caps. FRBNY Quarterly Review/Autumn 1985 along the lines of option pricing models, which assign probabilities to possible future interest rate paths and then average them.13 Rather than using this approach, we estimate the value of caps by calculating the extent to which they would have held down the interest costs of ARMs if they had been issued since 1970 (box). To represent the holding period of a mortgage, we use two horizons: eight and three years. The eight-year horizon approximates the average holding period of all mortgages, and as such is representative of the holding period for borrowers in the aggregate.14 We assume that most ARMs have an annual cap of two percentage points and a lifetime cap of five percentage points. These caps are among the most popular of the recFootnote 12, continued rate. The second element is a constant markup. The sum of these two is called the fully-indexed rate. The third element is the firstperiod discount, which reduces the fully-indexed rate for the first period of the mortgage only. The fully-indexed ARM rate less the first-period discount is called the initial rate. In the second period, the uncapped ARM rate has only two parts. It is the sum of the index rate as of the beginning of the period and the same markup as in the first period. 13See Randall J. Pozdena and Ben Iben, “ Pricing Mortgages: An Options Approach’1, Federal Reserve Bank of San Francisco Economic Review (Spring 1984), pages 39-55. 14See Frederick E. Balderston, Thrifts in Crisis (1985). ommended configurations proposed by the Federal National Mortgage Association. According to our calculations, the value of these caps for an eight-year horizon varied between zero and 1.8 percentage points, depending in part on the size of the first-period discount. For example, an ARM issued in 1970 without a discount would not have been affected at all by the presence of our caps. Thus, their value at that time was zero. In contrast, the rate on an ARM issued after 1971 would have been constrained not only by the annual caps but also by the lifetime cap. In these cases, the worth of the caps moved toward the high end of the range. As we expected, caps would have been less valuable to people with short horizons, e.g., three years, than to those with long horizons. The three-year and eight-year values differ mostly because lifetime caps were never binding over the first three years of an ARM during the 1970s. In general, our calculations indicate that bor rowers with short horizons face little likelihood that lifetime caps will ever come into play. To be sure, these individuals would value first-period discounts more highly than people with longer horizons since they amortize the discounts over fewer years. Nevertheless, we find that the combined value of caps and discount usually favors borrowers with longer horizons. Valuation of Caps, continued including prepayment of the principal, over the mortgage horizon assuming no caps; we call this the base present value. Second, we recalculate the present value imposing, individually and combined, caps of two per centage points each year and of five percentage points over the life of the mortgage. The differences between these values and the base present value measure the present values of the cost saving resulting from the respective caps. Expressing each difference as a percent of the face value of the loan converts the saving into the equivalent of closing points. Then calculating how much these points change the effective yield provides a measure of the effective value of the cap. For a borrower to accept an uncapped ARM instead of a capped ARM, its markup (over the base rate) would have to be lower by this effective value. (Equivalently, a larger first-period discount could be offered.) Since, in our simulations, the cap was tied to the initial rate rather than the fully-indexed rate, the value of a cap increases sharply as the discount increases. Since caps only have value when they lower the interest rate on the mortgage, their value depends on the course of future interest rates. Thus, an ex ante val 44 FRBNY Quarterly Review/Autumn 1985 uation is based on the steepness of the yield curve. When the yield curve rises sharply, reflecting a market expectation of high future interest rates, caps will be worth more. On the other hand, when the yield curve is downward-sloping, caps may turn into “floors” for bor rowers and could have a negative value. To capture the effect of the market yield curve, we divide the months of the simulations into three equal groups, ranked by value of the caps. (This ranking is similar to one based on the steepness of the yield curve at the time a hypothetical ARM was issued.) Then we average the values in each group. For each group we show, in the tables, for different first-period discounts, the effective value of the caps, singly and in combina tion. In examining the recent ex ante valuation, we use the relative steepness of the yield curve to select an appropriate value from the tables. For example, when the difference between the ten-year and one-year Treasury rates exceeds 1.5 percentage points, we use the average of the highest third as the value of the cap or caps. The average of the lowest third applies to yield curve differences of less than one-half percentage point. In cases near a boundary, we chose an average value of the two groups. Using our estimates of the value of caps, we next determine the extent to which the long-term expected ARM financing rate has been below the FRM rate. Financing rate: ARMs versus FRMs We evaluate the financing rates by comparing the initial period pricing of an ARM and the FRM rate with the corresponding points on the yield curve in the market for Treasury securities. Since the Treasury yield curve embodies only expectations of future rates and an interest rate risk premium in longer-term rates, sub tracting it from the yield curve implicit in the mortgage market shows the impact of the other factors that dis tinguish ARMs from FRMs . 15 Consider, for example, an 15Shiller, Campbell, and Shoenholtz, op. cit., show that long-term Treasury rates can be expressed as the sum of an interest rate risk premium and an arithmetic average of weighted expected future ARM without a discount or caps whose first-period rate is three percentage points above the one-year Treasury rate. If the FRM rate were only, say, two percentage points above a long-term Treasury rate, then the net effect of the distinguishing factors would make the long term expected financing rate of an FRM lower than that of an ARM by one percentage point (Chart 1) . 16 Footnote 15, continued short-term rates, where the weights sum to one. Expected rates receive less weight the further they are in the future. 16More precisely, in this case all expected ARM rates in the future are also three percentage points above expected future one-year Treasury rates, since the markup is constant. Thus, the expected long-term ARM rate exceeds the expected average one-year Treasury rate by three percentage points. This difference can be compared with the spread between the FRM and longer-term Treasury rate, in which the expected future short-term rates and interest rate risk premium are netted out. What is left over are the (p.46) Table 2 Value of Discounts and Caps in 1984 and 1985 In percentage points Eight-year horizon Size of discount* Quarter Yield cu rve t __________ Three-year horizon Effective value Effective value of discount}: of caps} Effective value Effective value of discount}: of ca p st 1984-1 .................................... .............2.0 1984-11 ................................................2.9 1984-111 .............2.5 1984-IV .............1.4 1.8 1.7 1.1 1.7 0.4 0.6 0.5 0.3 1.4 1.6 0.9 0.9 0.8 1.2 1.0 0.6 0.6 0.7 0.3 0.4 1985-1.. 1985-11 2.2 2.3 0.3 0.2 1.2 0.9 0.6 0.4 0.5 0.3 .............1.4 ............ .............0.9 ’ Discount is estimated as the excess of the sum of the one-year Treasury rate and 2.8 percentage points over the initial rate, as reported by the FHLMC. fDifference between the rates on ten-year Treasury notes and one-year Treasury bills. }:The effective values of the discount and caps are the consequential reductions of the effective yield of a mortgage over the stated horizon. Estimated by the authors using data from the Federal Home Loan Mortgage Corporation and the Federal Reserve Bulletin. Table 3 Evaluation of the Financing Rate of ARMs in 1984 and 1985 In percentage points ______________________________ Eight-year horizon Quarter \ Adjusted ARM rate less one-year Treasury rate’ FRM rate less ten-year Treasury rate Difference Three-year horizon Adjusted ARM rate FRM rate less one-year less three-year Treasury rate* Treasury rate Difference 1984-1 .............................. 1984-11 ............................ 1984-111 1984-IV 1.0 0.6 1.4 1.6 1.4 0.9 1.6 1.9 - 0 .4 - 0 .3 - 0 .2 - 0 .3 1.4 0.9 1.5 1.8 2.1 1.4 1.8 2.5 - 0 .7 - 0.5 - 0 .3 - 0 .7 1985-1.. 1985-11 1.3 .1.7 1.5 1.9 - 0 .2 - 0 .2 1.7 2.1 2.4 3.0 - 0 .7 - 0 .9 ’ Calculated as the constant ARM markup of 2.8 percentage points less the sum of the effective values of caps and discounts, shown in Table-2. FRBNY Quarterly Review/Autumn 1985 45 Evaluating the expected long-term ARM financing rate involves several steps. To take account of caps and the first-period discount, we add the present value of each to the face value of a mortgage and calculate the reduction of the effective yield over the holding period. We call this reduction the “ effective value” of the caps and first-period discount. By subtracting this effective value from the fully-indexed ARM rate in the first period, the net result, the “ adjusted” ARM rate, can be com pared with the one-year Treasury rate as previously discussed. We apply this approach beginning in 1984, which is the first year for which rates on a fairly homogeneous sample of ARMs are available. These data, compiled by the Federal Home Loan Mortgage Corporation (FHLMC), show the initial ARM rate. The difference between this rate and the one-year Treasury rate equals the markup less the first-period discount. To disentangle the firstperiod discount, we rely on two FHLMC surveys, taken at different times, indicating that the markup over the one-year Treasury rate for a typical ARM has been constant at 2.8 percentage points.17 On the basis of these survey results, we assume that all the variation in the initial ARM/one-year Treasury spread represents changes in the first-period discount. Since January 1984 this discount has varied between 0.9 and 2.9 percentage points, which, for an eight-year horizon, translates into a range of effective values between 0.2 and 0.6 per centage point (Table 2). Effective values of caps depend on the expectations of and the risks associated with future rates— both of which are embodied in the yield curve—and the dis count. Thus, we apply our estimated values to 1984 according to the slope of the yield curve and the size of the discount in each month, as described in the box. Caps were worth the most for ARMs issued in 1984-11 and the least in the second half of 1984 and 1985-11. Using the Treasury yield curve and our estimated values of the discount and caps, we now determine how attractive ARM pricing has been for the average holding period. For each quarter since the beginning of 1984 we calculate the adjusted ARM rate, as described above, and subtract from it the one-year Treasury rate. We then compare this difference with the spread between the FRM rate and ten-year Treasury rate.18 Table 3 shows that the two spreads were similar in every quarter, implying that the long-term expected financing rates of ARMs and FRMs were about the same. In other words, to the extent that individuals had the same expectations as the market, the average of expected ARM rates over the length of home ownership was close to the FRM rate. This has been the case when FRM rates were low, as in early 1984 and 1985, as well as when they were temporarily high, as in mid-1984. Even if ARMs do not appear to have been priced much below FRMs for the typical individual, ARMs might be favored by people with short horizons, e.g., an expected length of home ownership of three years, to avoid paying a long-term rate on a short-term loan. To evaluate the expected financing rate of an ARM for these borrowers, we compare the spread between the adjusted ARM rate and the one-year Treasury rate with the spread between the FRM rate and the three-year Treasury rate. In this comparison, the adjusted ARM/ one-year Treasury difference has varied between 0.3 percentage point and 0.9 percentage point less than the FRM/three-year Treasury rate difference since January 1984. While the differences may have been large enough to significantly affect these individuals’ demand 18The ten-year Treasury rate most closely matches the average holding period of a mortgage. Because the yield curve in the past several years has been essentially flat past a maturity of seven years, choosing other long-term rates does not significantly alter our results. C h a rt 1 ARM P ricing and the Treasury Market Yield Curve: An illu s tra tio n * P e rc e n t 15 14 13 12 11 Footnote 16, continued effects of the distinguishing characteristics of FRMs from ARMs. The long-term expected financing rate of an FRM would differ from that of an ARM by these effects as well as the interest rate risk premium. 17The first survey was taken in November 1984; see Federal Home Loan Mortgage Corporation, The Primary Mortgage Market (January 1985). The later survey, taken in February 1985, is unpublished. The one-year Treasury rate has gained in popularity as the index rate for ARMs over cost-of-fund indexes and by 1984 was used by about 90 percent of lenders surveyed. 46 FRBNY Quarterly Review/Autumn 1985 A < B-*-ARM is fa vorably priced. 10 A>B-*-ARM is not favorably priced. 10 15 20 Years to maturity * C onstru cte d by the authors. 25 30 Table 4 Spread Between the FRM and Ten-year Treasury Rates In percentage points 1970-78 average ..................................................................... 1979 1980 1981 1982 1983 1984 1.3 1.8 2.3 2.7 3.1 2.1 1.4 Federal Home Loan Mortgage Corporation and Board of Governors of the Federal Reserve System. for housing, this group includes somewhat less than a quarter of all homebuyers at a given time . 19 Thus, any resulting boost to aggregate housing demand is likely to have been relatively small. In sum, our estimates indicate that ARMs have gen erally not been priced significantly below FRMs. Inas much as our c a lc u la tio n s are based on several approximations, however, the precise estimates should not be taken literally. Nevertheless, the pricing of an ARM most likely has to be substantially more favorable than an FRM to persuade someone to purchase a house on the basis of the more risky financing rate. In this light, our results suggest that even if some of our approximations are not entirely correct, the alternatives are unlikely to be so different as to change the basic conclusion: ARMs do not seem to have been priced attractively enough to raise housing demand in the aggregate by a large amount. ARMs and the FRM rate ARMs may have still provided an indirect boost to housing by putting downward pressure on the FRM rate. Two arguments have been advanced along this line. First, to the extent that the FRM rate in the past con tained a premium to cover the risk associated with the im balanced portfolios of th rifts, the ARM-induced reduction of this risk might cut the premium . 20 Second, with ARMs having captured a growing share of new mortgages, the supply of FRMs in the secondary mort gage market may not have kept up with demand, especially after demand was bolstered by the devel opment of collateralized mortgage obligations in 1984.21 ’•John L. Goodman, op. cit. “ Robert Van Order, op. cit. 21See Joseph Hu and Judy Hustick, "Major Developments in Housing and Mortgage Finance", Bond Market Research, Salomon Brothers Inc. (January 1985). As a result, the price of FRMs may have been bid up, which reduced the FRM rate. Unfortunately, experience with ARMs has been too brief to distinguish their effect on the FRM rate from other influences. In fact, the FRM rate fell relative to other long-term rates over the past two years (Table 4). However, in 1982 the spread between them had wid ened to an unprecedented extent, most likely reflecting to some degree a jump in the FRM’s prepayment risk premium that ocurred when interest rates climbed to exceptionally high levels. The FRM rate subsequently declined relative to other rates at least in part because this risk premium fell along with the overall level of rates. The share of ARMs in newly issued mortgages Even though our calculations point to little impact of ARMs on housing demand, small differences in the perceived financing costs of ARMs and FRMs could still have a large effect on how people choose to finance a home. Because these two types of mortgages are so closely substitutable, the differences may greatly influ ence the choice between an ARM or an FRM once an individual has decided to purchase a home. Although it is very difficult to know at this point all the determinants of the share of ARMs in new mortgages, we investigate in this section two systematic factors that might tilt the financing choice: the distribution of risks surrounding the market’s expectations of future rates and the pattern of mortgage payments over time. In deciding whether to finance a home purchase with an ARM or an FRM, individuals presumably consider the risks surrounding market expectations of future rates. When interest rates look as if they will be rising, i.e., the yield curve slopes upward sufficiently, FRMs may be viewed as a better hedge than ARMs. Conversely, when rates look as if they will be falling, i.e., the yield curve is downward sloping, ARMs might be considered a good risk. From this perspective, then, the slope of the yield curve may indicate the predominant financing choice. Another factor that may influence the mode of home finance is the pattern of mortgage payments over time. One way to represent the different payment patterns of ARMs and FRMs is to use the difference between the FRM and the first-period ARM rates. The larger this spread, the lower the near-term payments of ARMs relative to those of FRMs and, thus, the more attractive ARMs may appear. Experience to date seems to support the roles played by these two factors. Since mid-1981 there have been two periods in which the share of ARMs in new mort gages has risen sharply—the first half of 1982 and the second half of 1983 through the first half of 1984 FRBNY Quarterly Review/Autumn 1985 47 Chart 2 Share of ARMs in Mortgages Closed Shading shows periods of increasing ARM popularity Percent The predicted values are from the estimated equation: ( t-statistics are shown in parentheses ) YTt = -2.333 + 1.130 IN ITG A P t - 0.366 SLOPEt-2 + 0.142 RFRMt-1 (-1.78) (8.17) (-3.40) (1.68) Sample period: January 1982 - July 1985 R2 = 0.65 D-W = 0.822 YT = loge(JARM) - loge(1-JARM) JARM = Share of ARMs in new single-family home mortgages closed INITG A P = Interest rate on FRMs less the initial interest rate on ARMs SLOPE = Yield on 20-year Treasury bills less the yield on one-year Treasury bills (constant maturity) RFRM = Interest rate on FRMs for new single-family homes The transformation of the predicted values of YT into the share is given by: A ^ JARMt = (1+eYTt)-i The dependent variable is so transformed to restrict its range to the interval [0.1]. Sources: Federal Home Loan Bank Board and Board of G o v e rn o rs of the Federal Reserve System . 48 FRBNY Quarterly Review/Autumn 1985 (Chart 2). During the first episode, the share peaked at 46 percent, and during the second period it reached 68 percent. Outside of these episodes—from the end of 1982 to the summer of 1983, and in late 1984 and early 1985—ARMs lost some of their popularity. In the first episode, the prim ary reason fo r the increased use of ARMs may have been related to the risks surrounding the yield curve. Over the first year or so since widespread use of ARMs was permitted in April 1981, the yield curve was downward sloping or fairly flat (Chart 3). People may have taken advantage of ARMs in the belief that the potential for future declines in interest rates made this form of financing a good risk. In contrast, the timing of ARM payments was probably not important since the initial ARM rate was not much different from the FRM rate during this period. The second surge in ARM popularity that began in the fall of 1983 may have been related to a widening spread between the FRM and initial ARM rates. In the spring of that year, first-period discounts became widely available and were more and more prevalent through the first half of 1984. These heavily advertised discounts may have reinforced people’s perceptions of the dif ferent payment streams associated with ARMs and FRMs. The yield curve was fairly steep during this period and, thus, was unlikely to be behind the growing share of ARMs in newly issued mortgages. However, the yield curve flattened substantially in the summer of last year and may have helped extend the popularity of ARMs through most of the remainder of 1984, despite a narrowing in the FRM/ARM spread. Finally, in almost all the periods when most people turned to FRM financing, neither the yield curve nor the FRM/ARM spread would have encouraged the wide spread use of ARMs; the yield curve was steep and the FRM/ARM rate difference small. Individuals with short horizons, however, would have chosen ARMs on the basis of the steep yield curve. Since the beginning of 1985, though, the FRM/ARM spread has begun to widen at the same time that the yield curve has remained very steep. So far, the share of ARMs has stayed around 50 percent, well below its previous peak. One factor that may be bolstering FRMs is that since the end of 1984 their rates have been close to their lowest level of this expansion. The long-term financing rate of a mortgage, thus, is perceived to be about as low as can be expected, thereby encouraging borrowers to lock in the FRM long-term financing rate.22 For confirmation that these systematic factors play a role in determining the mode of mortgage, we estimated “ S ee Freddie Mac Fleports, Federal H om e Loan M o rtg a g e C orporation (M ay 19 8 5), for a sim ilar analysis. Chart 3 Slope of the Treasury Market Yield Curve and Spread Between the FRM and Initial ARM Rates Shading shows periods of increasing ARM popularity Percent *D iffe re n c e between ten-year and one-year T reasury rates. Sources: Federal Home Loan Bank Board and Board of G overnors of the Federal Reserve System. several simple equations relating the share of ARMs to the yield curve, the FRM/ARM spread, and the level of the FRM rate. In the best of these equations, the explanatory variables, for the most part, were statisti cally significant and explained much of the variation in the share of ARMs (Chart 2). Conclusion Our analysis suggests that ARMs have not had a major effect on the demand for housing. We have shown that for most people the pricing of ARMs has been such that their expected* long-term financing rate may not have differed much from the FRM rate, assuming individuals have the same expectations of future rates as the market. For people who hold mortgages only a short time, the effects could be important, but this group tends to be less than one-quarter of all homebuyers at a given time. Nonetheless, small differences between ARMs and FRMs may have produced large swings in the mode of home finance, once the decision to purchase a house was made. We believe that this dual approach goes a long way in resolving the apparent paradox of the recent econometric findings that indicate little impact of ARMs on housing and the observed popularity of ARMs. Our results, to be sure, are based on short and limited experience with ARMs. The economy has not yet gone through a period of sharply rising interest rates while ARMs were widely available and familiar to most people. The impact of ARMs on housing demand might then be more pronounced than under recent financial market conditions. Judging from recent experience, however, our analysis also implies that ARMs have not significantly influenced the dynamics of the business cycle by altering the interest responsiveness of housing demand. Of course, ARMs may have other effects on the business cycle by making spendable income after mortgage payments, and thus consumption, more sensitive to interest rate changes. Nevertheless, since the long-term expected financing rate of ARMs seems to move broadly in line with the FRM rate, the aggregate demand for housing should continue to respond to interest rate movements as it has in the past. Carl J. Palash and Robert B. Stoddard FRBNY Quarterly Review/Autumn 1985 49 February-July 1985 Semiannual Report (This report was released to the Congress and to the press on September 6, 1985) Treasury and federal Reserve Foreign Exchange Operations During the period under review many observers of the foreign exchange markets were uncertain about the sustainability of the global economic expansion, now into its third year. The vigorous upswing in the United States had faltered in the third quarter of 1984, and market participants were anxious for evidence whether domestic demand would remain strong enough to sup port renewed increases in production and employment in 1985. Doubts developed about other countries’ ability to continue to expand should U.S. growth remain sub dued, since exports to the United States had been the major source of stimulus abroad. Meanwhile, inflation had decelerated in almost all of the industrial countries, but the scope for making further progress in the fight against inflation was seen as more limited at this stage of the business cycle. At the same time, market attention was focused on concerns about the imbalances in the structure of the current recovery— imbalances reflected in a large U.S. fiscal deficit, unprecedented disparities in the current account posi tions of the largest industrialized countries, interest rates at levels that appeared high relative to current inflation rates, and persistent unemployment problems abroad. With the major money and capital markets of the world increasingly integrated through progressive lib eralization of exchange controls and other regulations, A report by Sam Y. Cross, Executive Vice President, Federal Reserve Bank of New York and Manager of the Foreign Operations of the System Open Market Account. Officers of the Foreign Exchange Function, together with Richard F Alford, Elizabeth A. Goldstein, Thaddeus D. Russell, and Elisabeth S. Klebanoff contributed to its preparation. 50 FRBNY Quarterly Review/Autumn 1985 shifts in sentiment about these uncertainties were associated with sizable movements in dollar rates. During the six months February through July, the dollar briefly continued its four and one-half year climb, advancing strongly to hit record levels in the floating rate period. Thereafter it depreciated, at times quickly, to close the period much lower. The dollar’s continued rise: February to early-March The dollar was buoyed early in the period by an improving outlook for the U.S. economy and the impli cations for U.S. monetary policy. Data being published at the time pointed to a significant rebound in the fourth quarter that had been unanticipated just months before, and economic forecasters were beginning to present reassuring projections of moderate growth for 1985. An accelerating expansion of monetary aggregates was seen as limiting the scope for any further easing of U.S. monetary policy and might even suggest some tight ening. As a result, there was a perception in the market that the decline in U.S. interest rates, which had brought short-term deposit rates down more than three per centage points in about six months and was marked by two half percentage point cuts in Federal Reserve dis count rates, was not likely to continue. As this shift in expectations occurred, market rates for long-term as well as short-term instruments backed up somewhat during February and into early March. The economic outlook abroad was more guarded. The performance of many of the European economies had not been sufficient to dispel concerns about their longerterm growth potential. Industrial production statistics for the first quarter, while hard to interpret because of temporary disruptions associated either with labor dis putes or an unusually severe winter, pointed to declines in output in many large countries. Also, business opin ions and press commentary appeared to reflect a lack Chart 1 The Dollar against Selected Foreign Currencies Percent J A S O N 1984 D J F M A M J J 1985 Percentage change of weekly average bid rates for dollars from the average rate for the week of July 2-6, 1984. Figures calculated from New York noon quotations. of confidence in most countries that domestic demand could revive sufficiently to ensure a continued expansion should U.S. growth be subdued. Fiscal policies abroad were regarded as being almost universally restrictive, as the authorities sought further progress in achieving their medium-term goal of reducing fiscal deficits as a proportion of national income. Monetary policies were also generally restrained. Thus, few market observers thought that foreign central banks would welcome pressures emanating from either a renewed firming of interest rates in the United States or a continuing decline in their currencies to tighten monetary policy any more. Yet the impact on domestic prices of the progressive decline in these countries’ currencies against the dollar was showing through, at least in Germany where import prices were rising more quickly. Market participants therefore became wary of the possibility that the authorities there, as well as in other countries, might use intervention in an effort to stop the currency depreciations. The full range of these international issues had already been discussed at a G-5 meeting late in Jan uary. Moreover, the May 1983 Williamsburg agreement to undertake coordinated intervention as necessary was reaffirmed at that meeting and visible foreign exchange market operations had subsequently been undertaken by the authorities of several countries. Market partici- Chart 3 Real GNP and Real Domestic Demand Growth Chart 2 U.S. Interest Rates Percent 13--------------------------------------- 6l l J I I I I I I I I I I 1 _ L ,1 11 1 I I I 1 I I 1 I 1 F M A M J J 1985 In 1984 Percent 9 --------- Source: O rganization for Economic C ooperation and Developm ent, 1984. FRBNY Quarterly Review/Autumn 1985 51 pants perceived the central banks to be more willing to intervene than before. But they were uncertain about the circumstances in which the central banks would judge intervention to be appropriate. At the same time dealers remained impressed by the strength of demand for dollars in the exchange market. Enthusiasm spread about the degree of interest coming from abroad in the Treasury’s February refunding oper ations. Commercial entities were frequently seen as buyers of dollars, presumably to hedge future commit ments in light of the improving outlook for the dollar. As sentiment toward the dollar became increasingly bullish, the dollar rose through levels at which, in earlier months, some central banks had intervened and pre viously provided resistance. The dollar’s rise then gained momentum, markets became one-sided, and dollar rates moved quickly to successive highs against several European currencies. By February 26, the dollar had risen nearly 10 percent against major European cur rencies while rising 3 percent against the Japanese yen. At this point the dollar was at its highest level of the six-month period under review, trading around DM3.48 and $1.03 against the German mark and British pound, respectively. On three occasions during the first three weeks of February, the U.S. authorities intervened, selling a total of $242.6 million against marks, $48.8 million against yen, and $16.4 million against sterling to counter dis orderly market conditions in operations coordinated with foreign central banks. Between February 27 and March 1, the U.S. authorities sold another $257.4 million against marks in the New York market in a concerted intervention. These operations brought the total of U.S intervention sales of dollars, between the January 21 G-5 meeting and March 1 , to $659 million. As for the central banks of most other G-10 countries, they intervened much more heavily between February 27 and March 1 than before, selling dollars, buying German marks and other currencies, or doing both. For all G-10 countries as a group, the total of dollars sold during the five weeks between January 21 and March 1 was about $10 billion. This series of operations con stituted one of the biggest dollar interventions during the floating rate period. The sales of dollars by G-10 countries other than the United States was large enough to cause a sizable drop in their official foreign currency reserves. The decline: mid>March to end-July Even after the large interventions of late-February to early-March, the dollar traded close to its late February highs for about two weeks. But the intervention had resulted in an accum ulation of dollar-denom inated assets in private hands. Talk had begun to spread ear- 52 FRBNY Quarterly Review/Autumn 1985 Table 1 Federal Reserve Reciprocal Currency Arrangements In millions of dollars Institution Amount of facility July 31, 1985 Austrian National Bank ................................ National Bank of Belgium ........................... Bank of Canada ........................................... National Bank of Denmark ......................... Bank of England ........................................... Bank of France ............................................ German Federal B a n k ................................. Bank of Italy ................................................. Bank of Japan .............................................. Bank of Mexico ............................................ Netherlands Bank ........................................ Bank of Norway ............................................ Bank of Sweden ........................................... Swiss National Bank ................................... Bank for International Settlements: Swiss francs-dollars................................. Other authorized European currency-dollars..................................... 250 1,000 2,000 250 3,000 2,000 6,000 3,000 5,000 700 500 250 300 4,000 Total................................................................. 30,100 600 1,250 her that portfolio managers were gearing up to provide more currency diversification to customers’ portfolios, taking advantage of assets that appeared undervalued at current exchange rates and capitalizing on the pos sibility of future currency appreciation. Then, around mid-March, a more pessimistic reassessment of the outlook for the U.S. economy and a shift of view about interest rates began to weigh on the currency. By mid-March, a variety of statistics were indicating that economic activity in the United States was pro ceeding only at a relatively slow pace. W hile final demand remained buoyant, the demand for labor and growth of production in the manufacturing sector were much weaker than had been assumed in most forecasts earlier in the year. Market participants came to realize the extent that demand was being diverted away from U.S.-produced goods, thereby jeopardizing the sus tainability of economic expansion here. At the same time, signs of strain in U.S. financial markets became more prominent, raising the risk that financial as well as economic dislocations would inten sify. The failure of three secondary government secu rities dealers, though constituting a very small part of the market, imposed losses for a number of customers, including several local governments and thrift institu tions. The repercussions of these incidents revealed weaknesses in private deposit insurance systems and led to large deposit outflows from state-insured thrifts, particularly in Ohio, before the governor of that state temporarily closed the affected institutions. Pictures displayed prom inently by the media of queues of depositors unable to withdraw their funds heightened concern about the authorities’ ability to deal adequately with problem situations. Since difficulties had already been identified in energy, real estate, and agricultural portfolios, this weakness was perceived as having potentially far-reaching implications. Against this background, market participants adjusted their assessments of the outlook for U.S. monetary policy and interest rates. Dealers were sensitive to the implications of the imbalances in the economy for the industrial sector and the prospects for sustained growth. Money as measured by M1, though remaining well above target, was growing somewhat more slowly on a month-to-month basis. Inflation rates were still low, a renewed weakness in oil prices helped keep inflationary expectations at bay, and signs of Congressional action to reduce the fiscal deficit lent some relief to the bond market. Thus, most observers came to expect the Federal Reserve to give priority to supporting the economy and providing assistance to the domestic financial system. Market interest rates of all maturities started to decline in a trend that was to last about three months, while expectations developed that the Federal Reserve would announce a series of cuts in its discount rates. By midJune short-term interest rates had fallen two percentage points or more, with the Federal Reserve lowering its official rates just once— by half of a percentage point, effective May 20. Long-term rates also declined, but more slowly. As a result of these declines, most U.S. interest rates were below levels prevailing at the depth of the 1982 recession. As these developments began to unfold, the dollar fell substantially in the exchange markets. Many market participants were concerned for a time about the mag nitude of any drop in the dollar, if foreign investors tried to liquidate dollar assets accumulated during previous years. Indeed investors acted to protect the value of their portfolios, mostly by selling dollars in the forward market but also by shifting into assets denominated in other currencies. Commercial customers postponed dollar purchases in the expectation of being able to buy later at more attractive rates. Bank dealers and spec ulators on organized exchanges also sought to sell the dollar and to establish short positions. Under these circumstances the dollar moved lower. As it fell through levels at w hich re sista n ce had p re v io u s ly been expected, the pace of the decline quickened. From its peak in late-February to the middle of April, the dollar dropped 20 percent against sterling, 15 percent against the continental currencies, as well as 6 V2 and 4 percent against the Japanese yen and C anadian dollar, respectively. Late in April, however, the dollar firmed and then traded relatively steadily through the end of June. Market participants perceived that foreign investors had not liquidated their dollar holdings in large scale so that fears of an early and precipitous fall in the dollar faded. Instead, inflows of new funds were continuing, especially from Japan at the beginning of that country’s new fiscal year in April, as well as from countries suffering from serious inflation problems. Also, persistent strains in the U.S. financial sector were being well contained. Interest yields on dollar investments were still relatively attrac tive. The scope for hedging the currency risk should the dollar decline had been dem onstrated. And profits realized from earlier hedging operations increased the overall rate of return on dollar portfolios sufficiently to protect against even significant future declines in the dollar. In effect, the dollar retained its stature as the principal medium for investment. Meanwhile, the currencies that traditionally benefit from a shift of investor preference out of dollars, the German mark and Japanese yen, had appreciated rel FRBNY Quarterly Review/Autumn 1985 53 atively modestly as the dollar had declined. The U.S. economy had still outperformed those of most other industrialized countries and talk continued of a renewed acceleration of U.S. growth in the second half of 1985. The only currency to challenge the dollar as an invest ment alternative was pound sterling. With the outlook for economic growth in the United Kingdom brighter than for most other countries and interest rate levels there comparatively high, sterling-denominated assets pro vided an attractive outlet for investors reluctant to accept declines in yields elsewhere. Thus by the end of June, the dollar was trading above its mid-April lows against all currencies except sterling. Many market observers had supposed that the authorities abroad would have taken advantage of the decline in U.S interest rates that occurred during the spring to ease their own monetary policies. But in Germany and Japan the authorities appeared reluctant to cut short-term interest rates until they were more confident about the exchange market situation. In the other countries, the authorities were cautious about letting interest rates at home get too far out of line with those of their closest trading partners. To varying degrees, foreign central banks instead took advantage of the decline in the dollar to rebuild their foreign currency reserves. The authorities in several countries acquired sizable amounts of both dollars and German marks, currencies that could be used in future inter vention operations to support their own currencies. By the end of June the G-10 countries as a group had largely recovered the reserves lost in the early months of the year. In July the dollar resumed its decline. During the spring, the gap had continued between strong growth of U.S. dom estic demand and weak expansion of domestic production. As a result, the regular flow of economic statistics had presented conflicting signals. By early July, however, it again became clear that U.S. economic activity had not increased as much as most observers had expected. An acceleration of real GNP growth in the second quarter was more moderate than anticipated, and anecdotal information for July sug gested that the third quarter was getting off to no better a start. The mounting U.S. trade and current account deficits were increasingly perceived by market partici pants as a drag on the domestic economy. Noting an increase in protectionist pressures, they considered the possibility that the Administration might welcome a fur ther decline in the dollar to help restore external bal ance. At the same time, disappointment developed over the prospects for meaningful reduction of the fiscal deficit, as efforts in the Congress to adopt a compro mise budget resolution appeared to falter. During the month, interest rate developments tended to move in the dollar’s favor. In the United States, interest rates started to firm. Market participants here came to expect the Federal Reserve would not be more accommodative until it could assess more fully the implications of the drop in interest rates that had already occurred and of a renewed acceleration in M1 growth. In Europe, interest rates began to ease more rapidly. The central bank in Germany began to provide liquidity at progressively lower interest rates and, at least for a time, central banks in other continental countries moved in a similar direction. Thus, interest differentials actually moved in favor of the dollar during the month. Nonetheless, sentiment toward the dollar had become cautious. Market professionals had already begun to set up positions in anticipation that the dollar might resume its decline. Thus, when others came into the market to sell, dollar rates moved down through the end of the month, dropping well below the lows of mid-April. Ster ling continued to lead the rise in foreign currencies against the dollar. After mid-July, however, when a Table 2 Drawings and Repayments by the Argentine Central Bank Under Special Swap Arrangements with the U.S. Treasury In millions of dollars; drawings ( + ) or repayments ( - ) Drawings on the United States Treasury $500 million .............................................. $150 million .............................................. Data are on a value-date basis. ’ Not applicable. 54 FRBNY Quarterly Review/Autumn 1985 Outstanding September 31, 1984 1984-IV 1985-1 1985-11 Outstanding July 31, 1985 * + 500 -2 3 0 -2 7 0 -0- -0- + 75 + 68 + 143 realignment within the European Monetary System (EMS) drew attention to the mark’s potential for reval uation in that arrangement, the German currency also began to strengthen more rapidly than before. During the entire February-July period under review, the dollar had fallen on balance 20 percent against sterling to $1.4135, 12 percent against the mark to DM2.7850 and by approximately similar magnitudes relative to most other continental currencies, and by 8 percent against the Japanese yen to ¥236. Meanwhile, during late June and July, progress was being made in some of the largest Latin American countries to deal with the serious imbalances in their economies. In Argentina, the government came to an agreement with the International Monetary Fund (IMF) on a stabilization program that entailed currency and wage/price reform designed to brake the country’s rap idly accelerating inflation. Upon completion of an agreement by the IMF to provide a standby, the U.S. Treasury and 11 other monetary authorities acted to facilitate the provision of a $483 million bridge financing facility for Argentina, of which the U.S. portion was $150 million. Argentina made two drawings of roughly equal size on this facility, on June 19 and on June 24, for a total of $460 million. The Treasury’s portion of these drawings was $143 million. Argentina is scheduled to repay the drawings in two installments after the period. In Mexico, the government tightened fiscal policy, lib eralized trade policy, and made major changes in the structure of its exchange market. These actions were undertaken in order to align Mexico’s cost and price structure more closely with world markets and aid in bringing inflation down to targeted levels. In the period February through July, the Federal Reserve and the Exchange Stabilization Fund (ESF) realized no profits or losses from exchange transactions. As of July 31, cumulative bookkeeping or valuation losses on outstanding foreign currency balances were $871 million for the Federal Reserve and $578 million for the Treasury’s Exchange Stabilization Fund. These valuation losses represent the decrease in the dollar value of outstanding currency assets valued at end-ofperiod exchange rates, compared with the rates pre vailing at the time the foreign currencies were acquired. The Federal Reserve and the ESF invest foreign currency balances acquired in the market as a result of their foreign operations in a variety of instruments that yield market-related rates of return and that have a high degree of quality and liquidity. Under the authority pro vided by the Monetary Control Act of 1980, the Federal Reserve had invested $1,009.2 million equivalent of its foreign currency holdings in securities issued by foreign governments as of July 31. In addition, the Treasury held the equivalent of $1,756.0 million in such securities as of the end of July. Chart 5 Changes in Currency Reserves of G-10 Countries Excluding United States Billions of U.S. dollars equivalent mam Table 3 Net Profits ( + ) or Losses ( - ) on United States Treasury and Federal Reserve Current Foreign Exchange Operations In millions of dollars Period February 1 - July 31 .. Valuation profits and losses on outstanding assets and liabilities as of July 31, 1985 Federal Reserve United States Treasury Exchange Stabilization Fund -0- -0- i— Jan i Feb i Mar i Apr 1985 i i May i Jun J Jul Foreign currency reserves shown in this and the following charts are drawn from IMF data published in International Financial Statistics. -871.1 Data are on a value-date basis. -5 7 8 .3 Adjustm ents fo r gold and fo re ig n exchange swaps against European cu rre n cy units done with the European M onetary Fund are incorporated. FRBNY Quarterly Review/Autumn 1985 55 European currencies Coming into the six-month period, progress appeared to stall in resolving the economic problems facing Euro pean countries. During the months of severe winter weather, growth in several countries slowed, unem ployment in some continued to drift upward, and a deceleration in inflation petered out. At the same time the trend toward greater convergence of economic performances started to dissipate, notwithstanding the fact that governments in almost all of these countries continued to be committed to common goals for eco nomic policy: reducing government deficits and con taining inflation. Under these circumstances, there were some adjustments among the relationships of all Euro pean currencies as they declined and then rose against the dollar. Early in the period, with the dollar strengthening across the board, the continental currencies as a group fell about 10 percent. The Swiss franc dropped to SF2.9405, the lowest level in more than 10 years, and the German mark posted a low for the floating-rate period at DM3.4780. The Dutch guilder, the French and Belgian francs, and the Italian lira dropped to record lows of NG3.9430, FF10.6300, BF69.90, and LIT2167, respectively. Sterling, which had been the target of especially heavy selling pressure just before the period, declined somewhat more slowly against the dollar during February. Nevertheless, by February 26 it had declined nearly 9 percent and also recorded a record low of $1.0370. Meanwhile, authorities in Germany and the United Kingdom were concerned that inflation was picking up as a result, at least in part, of the impact on import prices of the continuing strength of the dollar. In the United Kingdom, inflationary expectations were also stimulated by concerns over the priorities of the gov ernment’s economic policy and above-target growth of money. But the British authorities had acted to address these concerns prior to the period by permitting an abrupt and sharp increase in short-term interest rates. In Germany, where the pressures were far less acute, market rates also tended to firm. But market participants perceived the German authorities to be resisting the rise out of concern that significant increases in interest rates were not appropriate to the domestic economic situation. These developments had disappointing implications for other countries that had been maintaining favorable interest rate differentials relative to Germany. The cen tral banks in France, Italy, and Belgium, for example, saw the opportunity for them to lower interest rates in response to earlier improvements in their price per formance as quickly slipping away. Following the G-5 meeting in January most European central banks participated in the coordinated interven 56 FRBNY Quarterly Review/Autumn 1985 tions that took place through early March. All of those participating sold dollars, at times in sizable amounts. Some supplemented their dollar sales with purchases of marks and a couple of other currencies, either against dollars or their own currencies. From mid-March, when the dollar began to decline, to end-June, sterling was the currency that rebounded most strongly to lead the rise in European currencies against the dollar. The Swiss franc also benefited more than many others, while the German mark was not particularly buoyant. This pattern of exchange rate changes surprised market observers who had anticipated that, once the dollar started to fall, the mark would reassert itself as the principal alternative for investment. But as it turned out, the currencies to benefit most from the dollar’s ini tial decline were, for the most part, those with assets yielding relatively high interest rates. Foreign capital was drawn into sterling, enticed by high yields on gilts and other fixed income securities as well as the breadth and liquidity of London’s financial markets. Residents in high Chart 6 Selected Foreign Currencies against the Dollar Percent 35-------- 5L I J I I I F I I I I I M A I I I I I I I I M I I I J I I I I I J A 1985 Percentage change of weekly average bid rates for selected foreign currencies from the noon rates on February 28, 1985. Figures calculated from New York noon quotations. II interest rate countries borrowed abroad where the cost of funds was lower to finance trade and domestic expenditures. The Swiss franc firmed against many other currencies, even though Swiss interest rates remained relatively low, because the impression spread in the markets that monetary policy in Switzerland was not likely to be eased. In Germany, interest rates were also lower than in most other countries, and economic indicators for the first quarter were being interpreted in the market as disappointing. Expectations developed that the Bundesbank would cut interest rates as soon as exchange market conditions permitted and U.S. interest rates declined. Although the upward pressure on European interest rates subsided as the dollar declined during the spring, the European monetary authorities were slower to reduce interest rates than many market observers had expected. In the United Kingdom, the authorities were intent on reassuring markets of their commitment to strict financial policies. A cautious budget, presented in March, called for both a drop in the public sector borrowing require ment and reductions of growth targets for Britain’s two monetary target variables, MO and M3. As interest rates in the United States declined and capital inflows into sterling exerted upward pressure on the pound, the Bank of England allowed interest rates to ease some what. But the authorities were perceived as acting to slow the decline—an approach that appeared reason able as long as the economic outlook for the United Kingdom was more optim istic than for most other countries. By late June, short-term interest rates were still above 12 percent and differentials v/'s-^-v/'s dollar interest rates were even wider than they had been early in February. In Germany, also, the Bundesbank did not judge the domestic situation as warranting a change in the course of monetary policy. The central bank saw the underlying trend of economic activity still pointing upward. Central bank money stock was growing close to the top of its target path, buoyed by an acceleration of domestic credit growth early in the year. The public sector in particular was tem porarily having an expansionary impact on monetary growth. And by late spring a public debate had emerged over accelerating proposed tax cuts. The Bundesbank did not wish to suggest that an easing of policy was appropriate by announcing reduc tions of its official rates. But it was willing to provide sufficient liquidity to the banking system mainly through repurchase agreements. These operations reduced banks’ use of Lombard credit and guided day-to-day money rates cautiously lower. By the end of June, threemonth money rates had eased 75 basis points from Chart 7 Selected Interest Rates Chart 8 Three-month m aturities* United Kingdom Percent Movements in exchange rate and official foreign currency reserves Dollars per pound 1 .5 0 --------------- Billions of dollars 1.5 - 0 .5 1984 1985 Exchange rates shown in this and the following charts are weekly averages of noon bid rates for dollars in New York. 1984 ♦W eekly averages of daily rates. 1985 * Foreign exchange reserves for the United Kingdom and other members of the EMS, including Germany, incorporate adjustments for gold and foreign exchange swaps against European currency units done with the European Monetary Fund. FRBNY Quarterly Review/Autumn 1985 57 levels of end-February, less than half the decline for comparable rates in the United States. The relative stability of interest rates in Germany was a factor limiting the scope for interest rate declines in other European countries. The authorities there had accepted that domestic interest rates would remain considerably higher than those in Germany because inflation rates were higher and current account positions were not as strong. Yet their currencies were being buoyed relative to the mark by the inflow of interestsensitive capital. Under the circumstances, these central banks also looked to relatively subtle techniques to ease money-market rates gradually, so as not to suggest that a change in policy was underway. The Bank of France, for example, lowered its money market intervention rate, acting cautiously by moving in several small steps. In this way, short-term interest rates in France declined somewhat more than in Germany. A more substantial change in technique occurred in Belgium where the National Bank decided to adopt a more flexible and market related practice for fixing the discount rate. Henceforth the discount rate was to be linked to the rate on three-month Treasury certificates. As a result, a decline that had already occurred in market rates was acknowledged and rates continued to ease modestly through the end of June. Against this background, the authorities in many European countries also chose to respond to the favorable exchange market environment for their cur rencies by acquiring foreign currency reserves. During the second quarter a number of central banks were active buyers of dollars either in the market or from customers. They also purchased substantial amounts of other currencies, especially the German mark, because it is a currency frequently used for intervention within the EMS and is of increasing importance in the reserve holdings of other European countries. As a result of these operations, many countries restored the reserves lost during their intervention operations in late January through early March. France and Italy had among the largest increases in reserves. Germany’s increase was the greatest, even though it refrained from intervening for much of the period. Meanwhile, the Italian lira had broken stride with the other European currencies. During February it had risen against the dollar more slowly than the others. As a result, it had moved from the top to the bottom of the narrow EMS band between early February and midMarch and then traded consistently about 11/2 percent below the bottom-most currency in the narrow band during the second quarter. Fiscal policy in Italy had been expansionary, with the government deficit expected to grow to 17 percent of GDP in 1985. Moreover, Italy’s inflation remained high relative to that of other countries and successive increases in wage settlements eroded the country’s competitiveness all the more. Accordingly, the current account had deteriorated, with imports of capital goods quickening. Under these circumstances, market participants came to anticipate that the Italian authorities might welcome a decline in their currency. Sentiment toward the lira was briefly buoyed in May and June when the government’s position strengthened with a defeat of a referendum reinstating wage index- Chart 10 Chart 9 France Germany Movements in exchange rate and official foreign currency reserves Movements in exchange rate and official foreign currency reserves Marks per dollar Billions of dollars ------ Francs per dollar Billions of dollars 2.0 1.0 0 - 1.0 - 2.0 - 3 .0 - 4 .0 J A S O 1984 N D J F M A M 1985 J See notes on Chart 8. 58 FRBNY Quarterly Review/Autumn 1985 J 1984 See notes on Chart 8. 1985 ation and a smooth transition to a new presidency. But by July the lira had resumed its slide toward its lower EMS limit. This depreciation helped to offset the com petitive disadvantage resulting from accumulated infla tion differentials but removed room for movement of the exchange rate within the wide band available to the lira in the EMS arrangement. The Italian authorities there fore decided to seek a realignment of the lira’s central rates. Thus, after the lira dropped to its existing lower limit in hectic trading on Friday, July 19, the authorities closed the foreign exchange markets in Italy after the fixing. That weekend the EMS countries agreed to a realignment that took the form of a 7.8 percent deval uation of the lira’s bilateral central rates against all other active EMS members. As a result, the lira’s European currency unit central rate fell by 7.7 percent while the others rose by 0.15 percent. The July realignment of the EMS served to focus market attention on the risks of further adjustments in the exchange rate relationships among European cur rencies. Market operators began to hedge their bor rowings in low interest rate currencies and their investments in high interest rate currencies. The mon etary authorities in countries like France and Belgium found the scope for letting interest rates ease or for adding to official reserves more circumscribed than before. At the same time the Bundesbank found that the exchange rate environment, together with a reaffirmation of the governm ent’s policy of fiscal consolidation, afforded an opportunity to let short-term interest rates decline more quickly. A similar development occurred in the Netherlands. About the same time in July sentiment toward sterling began to soften as well. The pound had risen progres sively against the mark to levels that brought into question B ritain’s com petitive position vis-a-vis its European trading partners. Moreover, the earlier opti mistic assessment of the country’s economic prospects gave way to a more guarded outlook in the face of a weakening flow of new orders and a flattening of output growth. Market participants came therefore to expect the Bank of England to permit a more rapid decline in interest rates, even if the pound were to weaken as a consequence. Indeed, during the month, money market rates in London declined toward the 11 percent level and favorable interest rate differentials relative to the dollar narrowed by about one and one-half percentage points. In response, sterling gave up some of its gains vis-a-vis the mark late in the month. Thus, the decline in the dollar in July came to be reflected in a somewhat more rapid rise in the German mark than before. Even so, at the end of the six-month period under review, the pound had still risen from the February lows against the dollar by more than the other Chart 11 Movements in exchange rate and official foreign currency reserves Lira per dollar 1500 Billions of dollars 1.5 1600 1700 1800 1900 2000 Exchange rate ------Scale 2100 J A S O 1984 N D M A M - 2.0 1985 See notes on Chart 8. Chart 12 Percent D eviation of Selected EMS C urrencies from th e ir B ilateral C entral Rates* EMS Percent realignm ent *W e e kly averages of daily 9 a.m. rates. "^"The Italian lira may fluctuate i 6 percent from its central rate with other participating currencies. FRBNY Quarterly Review/Autumn 1985 59 European currencies. It closed the period up 38 percent from the end-February lows at $1.4350. The mark rose 25 percent during the same period to DM2.7800, with the Swiss franc and most EMS currencies moving roughly in line with the mark. The lira rose 18 percent to LIT1872. Japanese yen The yen generally moved in line with European curren cies against the dollar during the six-month period, but its fluctuations were narrower. As the period opened, market sentiment toward the yen was relatively positive. An annualized 9 percent rise in GNP in the fourth quarter of 1984 and optimistic projections for calendar 1985 compared favorably with the experience and out look of other countries. Inflation remained low, with the effect of the yen’s depreciation against the dollar offset by its rise against other currencies and by the weakness of world commodity prices, particularly petroleum. Japan’s current account surplus had grown to a record $35 billion in 1984. Thus the yen did not fall as rapidly against the dollar as the European currencies during February. Japanese fiscal policy continued to be one of grad ually reducing the government’s fiscal deficit as a pro portion of GNP. The Bank of Japan maintained its accommodative monetary stance, but the central bank refrained from reducing its official lending rates, citing as its main reason the need to support the yen in the exchange markets. After March the yen did not rise as rapidly as other currencies against the dollar. Attention was often focused on Japan’s huge long-term capital outflows— which had reached $50 billion in 1984—as a major potential source of unpredictable pressure against the yen. At times during the period, the yen’s performance in the exchange market— as well as credit market developments in both Japan and the United States—was influenced temporarily by reports and rumors about possible changes in rules or preferences governing Japanese investment abroad. In any case, the yen did not benefit, as did the European currencies, from a favorable shift of capital flows late in the period under review. Long-term capital outflows, as measured in Japanese net purchases of foreign bonds, actually grew larger to set new records in June and July. But since a greater proportion of the outward investment by Japa nese residents than before was thought to be hedged through forward foreign exchange transactions and short-term dollar borrowings, the resulting pressures against the yen were substantially mitigated. Rising foreign protectionist threats against Japan, and demands that the Japanese government step up its actions to reduce the trade imbalance, also attracted 60 FRBNY Quarterly Review/Autumn 1985 Chart 13 Japan Movements in exchange rate and official foreign currency reserves Yen per dollar Billions of dollars 235------------------------------------- -------------------------------------------- 0.8 Exchange rate 1984 1985 See exchange rate footnote on Chart 8. attention in exchange markets at times as a potentially negative background factor for the yen. Generally, however, such pressures did not have im m ediate exchange-rate influences. Announcements in April and June of new Japanese government programs to open domestic markets by reduced tariffs, liberalized invest ment rules, and adm inistrative reform s had little apparent impact on the yen rate at the time. By the end of the period, Japanese foreign currency reserves had risen by almost $1.2 billion to $2.38 billion, largely reflecting interest earnings. Canadian dollar The Canadian dollar, like other currencies, weakened considerably against the U.S. dollar early in the period. The rise in U.S. interest rates during January and Feb ruary fanned renewed debate over priorities for mone tary and fiscal policies in Canada. Inflation in Canada had stabilized under 4 percent on a year-on-year basis but the unemployment rate had recently moved back over 11 percent. Market participants, noting that Can ada’s traditional interest rate advantage had dwindled to about one percentage point by early February, questioned the willingness of Canadian authorities to permit increases in interest rates comparable to those in the United States. Moreover, uncertainty developed as to whether Canada’s newly elected government would deal decisively with its plan to reduce the budget deficit and improve the investment climate. At the same time unease developed surrounding potential capital outflows related to the acquisition by Canadians of foreign-owned assets in the petroleum sector. Chart 14 Chart 15 Canada Interest Rates in Canada and the Eurodollar Market Movements in exchange rate, o fficia l foreign cu rre n cy reserves, and in te re st d iffe re n tia l Canadian dollars per dollar Three-month m a tu ritie s* Billions of dollars 2.0 Exchange rate y T A •*------ bcaie Percent 14 1.5 Foreign currency 1.0 ^ S ca le ----- ► 0.5 B 0 - 0 .5 LI I I I I I IV | | I I I - 1.0 Percent 7 2 2 J A S O 1984 N D J F M A M 1985 *W eekly averages of daily rates. 1 0. 0 J A S O 1984 N D J F M A M 1985 J J See exchange rate footnote on Chart 8. ♦C anadian finance paper minus Eurodollars. Weekly average of daily rates. Against this background, sentim ent toward the Canadian dollar deteriorated sharply. Speculative selling and an adverse shift in commercial leads and lags put pressure on the exchange rate which fell to an all-time low of Can.$1.4070 ($0.7107) early in March, a decline of 6 percent from the end of January. The authorities intervened heavily to moderate the decline, financing their dollar sales by drawing on the government’s credit lines with commercial banks and borrowing in the Eurodollar market. Moreover, the Bank of Canada allowed interest rates to rise more sharply than U.S. rates, and the currency’s interest rate advantage wid ened to 21/2 percentage points. These developments helped to convince market par ticipants that the authorities’ approach to the exchange rate had not been changed. In addition, the Canadian government announced plans for tax increases and expenditure cuts to reduce the fiscal deficit together with legislation to remove impediments to foreign investment in Canada, thereby reducing uncertainty further. More over, a strong external performance, signs of a pickup in the domestic economy, and low wage settlements provided a more encouraging outlook for the currency. Thus, the Canadian dollar recovered after mid-March most of the ground it lost earlier in the period to close at Can.$1.3539 ($0.7386), down only 2 percent on bal ance over the six months. Under these circumstances, interest differentials eased back to fluctuate around 1 1/ 2 percentage points over the remainder of the period. The Bank of Canada made net dollar purchases as its currency rose, which it used to repay debt on its com mercial bank credit lines and bolster reserves. In addi tion, a further U.S. dollar borrowing in the U.S. market served to boost the level of foreign exchange reserves. By the end of July, foreign exchange reserves were up $498 million over the period under review at $2.1 billion. Selected Latin American currencies During the six months under review, two major Latin American countries, Mexico and Argentina, introduced new economic packages that included, among other measures, reforms to their respective foreign exchange systems. In the case of Mexico, this package was designed to get its stabilization efforts of the past three years back on track. In the case of Argentina, the task was to embark on major reforms to reverse long fes tering economic imbalances that were being reflected in spiraling inflation rates. FRBNY Quarterly Review/Autumn 1985 61 Mexico Mexico had posted a significant improvement in its trade account, which had swung from a deep deficit into sur plus in 1983 and 1984. However, the surplus had sub sequently narrowed. During the first four months of this year, the weakening of Mexico’s external position was being accentuated by a nearly 10 percent fall in total exports. Oil shipments dropped in the face of weakening prices elsewhere, the competitiveness of non-oil exports declined with a real appreciation of the “ controlled” exchange rate, and the pressures of increasing internal demand deflected production to the home market. Under these circumstances, Mexico’s current account surplus for all of 1985 was also expected to diminish, notwith standing the reduction of interest payments stemming from declining interest rates. Meanwhile, Mexico’s fiscal deficit through June rose to well above target levels. The budget overrun reflected the lower-than-anticipated oil revenues and increased gov ernment spending resulting partly from higher-thanexpected inflation and greater internal interest payments. In response to these pressures, beginning in late May the discounts widened between Mexico’s “ controlled” exchange rate for licensed transactions and the two free market rates—the internal “ free” rate and the “ superfre e ” rate across the M exican border. Thus, the improvement in the foreign exchange position of the Mexican peso, which had occurred in late March and in April following announcement of new understandings with the IMF on 1985 economic policies and the signing 62 FRBNY Quarterly Review/Autumn 1985 of the first phase of Mexico’s multi-year rescheduling, quickly dissipated. By late spring the external market was subject to recurring rumors of an impending peso devaluation, an increase in the daily rate by which the authorities adjusted the crawling “ controlled” rate, and cuts in oil export prices. By mid-July, the gaps between exchange rates for the peso were increasingly large. Exporters had the incentive to delay or divert revenues required to be converted in the “ controlled” market to either the domestic “ free” market or the external, “ super-free” market. Also, the volume of trading in the internal “free” market diminished substantially. Thus, the widening gap of peso rates was a source of growing concern to the authorities. To deal with this situation, the Mexican authorities adopted a series of measures, starting in mid-June. Under M exico’s procedures for licensing im ports, exporters were granted certificates of importation rights (called “ DIMEX” ), permitting them to import without license a range of raw materials and inputs to make their operations more efficient. Effective June 28, Mex ican banks were allowed to operate in the foreign exchange market at the “ super-free” rate by establishing trading houses designed for this purpose. After the Mexican banks were able to participate in the “ superfree” market via their trading houses, they became major intermediaries in that market. Then, on July 11, the Mexican banks, supported by the monetary author ities, decided to stop trading at the internal “free” rate. As a result transactions were switched from the “ free” market, where the peso was trading at 247.3 pesos per dollar the day before, to the “ super-free” market, where the peso was at 312.0 pesos per dollar before the announcement of this change. This switch constituted a 26 percent devaluation for transactions not eligible for the “ controlled” rate. Then on July 25, the Mexican government announced additional economic reforms including: • A 17 percent devaluation of the ‘ ‘c o n tro lle d ” exchange rate, from 232 to 279 pesos per dollar. • The introduction of a “ regulated float” to replace the earlier crawling system involving a fixed, daily slide of the peso against the dollar for the “ controlled” market. • Elimination of import permits on goods accounting for about 37 percent of its imports, thereby making a total of over 60 percent of Mexican imports sub ject to tariffs rather than non-tariff barriers, and a further enlargement of the “ DIMEX” arrangements. • A cut in c u rre n t g o vernm en t e x p e n d itu re s , Chart 17 Mexican In fla tio n Rate Percent 1 20 ------------------------------------------------------ 1982 1983 1984 1985 Sources: International Monetary Fund, International Financial S ta tis tic s : and Banco de Mexico, Informe Anual. amounting to 150 billion Mexican pesos during 1985, that entailed a 20 percent cut in budgeted expenditures on goods, the elimination of several highly visible government positions, and major cutbacks in expenditures by public enterprises. The purpose of these reforms was twofold. First, they were expected to relieve demand pressures in the economy coming from the public sector. Second, they were intended to improve competitiveness by adjusting the exchange rate and by opening the domestic market to lower-priced imports for raw materials, intermediate products, and capital goods. During the period between the announcement of the abolition of the internal “ free” market and the rest of the economic reforms, the peso weakened sharply as Mexican residents rushed to buy dollars in anticipation of a further devaluation. By July 24, the market rate in Mexico and abroad had fallen a further 20 percent to 374 pesos per dollar, and the discount relative to the “ controlled” rate widened to more than 60 percent. But by the end of July, the peso recovered to 354.50 pesos per dollar, and the discount from the “ controlled” rate narrowed to about 27 percent. Argentina In Argentina a newly constituted democratic government had been attempting to grapple with a debilitating wage/ price spiral without jeopardizing promised increases in real incomes. But the domestic economy was in severe disequilibrium. The central bank had monetized years of oversized fiscal deficits. It found that, with public sector wage increases and fiscal policy stimulating demand, efforts to restrict excessive bank lending through interest rate ceilings and credit allocation schemes led to a diversion of financing to an informal inter-company market. Argentine officials had repeatedly spoken of the need for programs to stabilize the economy over time by tightening monetary and fiscal policies. As recently as December 1984, Argentina had announced a 15-month standby arrangement with the IMF. But the country was from the start not in compliance with the standby pro visions and the rise in Argentina’s inflation rate con tinued to accelerate. In the process, the strategy of gradual adjustment had lost credibility. By early 1985 the internal chaos wrought by an economy reeling toward hyperinflation provoked political demands for a new approach that promised quicker results, even if the approach involved immediate sacrifice. Thus, in March President Alfonsin, with a new economic team, began to adopt a series of new measures to achieve rapid adjustment and a radical restructuring of the economy. First, regulated deposit rates were raised to levels comparable to the monthly inflation rate. Interest rates were deregulated on some bank liabilities to attract funds back into the banking system where the authorities could exert more control on credit creation. Public utilities also raised prices significantly to increase revenue. On June 11, the government announced an 18 per cent devaluation of the Argentine peso in the official market. Previously, the government had implemented “ mini-devaluations” rarely exceeding 4 percent, and averaging about 1 percent per day to adjust for the inflation differentials between Argentina and other countries. Following this action, and amid rumors of dramatic economic measures, the premium which Argentine residents had to pay for dollars in the parallel market widened to 35 percent. Then on June 14, President Alfonsin announced a package of bold economic reforms, centering on a further, substantial cut in the fiscal deficit and a pledge to stop monetizing the deficit. The deficit, which had fluctuated in the range of 10 to 12 percent of GDP since the end of 1983, was to be slashed to only 2.5 percent for the second half of this year. In support of this plan, price and wage ceilings were fro ze n — actions described as interim steps toward eliminating the country’s price and wage indexation system that FRBNY Quarterly Review/Autumn 1985 63 Chart 19 Chart 18 Argentine Inflation Rate Argentine Currency Dollar exchange rate for p e s o * Dollar exchange rate for a u s t r a l * Percent 1250 — 10 00 750 500 250 L...... 1982 *O ne austral = 1,000 pesos (weekly observations). was perpetuating Argentina’s inflation problem. In addition, currency reform was instituted to replace the Argentine peso with a new currency, the austral, at a rate of 1000 pesos to 1 austral. Effective June 16, the austral was given a fixed parity of 80 austral cents to the U.S. dollar. On the basis of these measures the government was able to shore up Argentina’s external financing position and reduce cash flow problems. It completed negotia tions for reactivating the IMF program, which was approved on August 9. It also took steps to reduce interest arrears on public sector debt, using funds from 64 FRBNY Quarterly Review/Autumn 1985 I 1983 I 1984 I 1985 Sources: International Monetary Fund, International Financial Statistics. official reserves and drawing upon a multilateral bridge financing facility backed by the monetary authorities of the United States and 11 other participating countries. The government’s actions also set the stage for com pletion of a rescheduling agreement and a new lending program with commercial banks. The announcement of the government’s adjustment program was generally well received in Argentina. In the exchange market, too, the Argentine currency appeared to have gotten a steadier footing by late July. Capital inflows began to materialize, taking the form at least in part of a reversal of commercial leads and lags. NEW PUBLICATION The Federal Reserve Bank of New York has issued a revised and expanded version of its booklet, Open Market Operations. The 48-page booklet by Paul Meek gives an insider's view of the mechanics of open market transactions and the implementation of monetary policy. The booklet evolved from four earlier editions by Mr. Meek. It is now directed at undergraduate students of economics, participants in the financial markets, and the general public. Mr. Meek retired earlier this year as vice president and monetary adviser in the New York Fed’s open market operations area. Topics covered in the booklet include: • How the New York Fed carries out open market operations on behalf of the Federal Reserve System by purchasing securities to supply reserves to the banking system and selling securities to withdraw reserves; • The significance of Federal Reserve float, Treasury cash balances, and currency in circulation in managing bank reserves; and • The trading desk’s daily agenda, as well as Federal Open Market Com mittee meetings with primary dealers and conferences with representatives of the Treasury and Board. Single copies of Open Market Operations are available free from the Public Information Department, Federal Reserve Bank of New York, 33 Liberty Street, New York, N.Y. 10045. Reasonable quantities are available upon request. Subscriptions to the Quarterly Review (ISSN 0147-6580) are free. Multiple copies in reasonable quantities are available to selected organizations for educational purposes. Write to Public Information Department, 33 Liberty Street, New York, N.Y 10045 (212-791-5000). Single and multiple copies for United States and for other Western Hemisphere subscribers are sent via third- and fourth-class mail, respec tively. All copies for Eastern Hemisphere subscribers are airlifted to Amsterdam, from where they are forwarded via surface mail. Multiple-copy subscriptions are packaged in envelopes containing no more than ten copies each. Quarterly Review subscribers also receive the Bank’s Annual Report. Quarterly Review articles may be reproduced for educational or training purposes only, providing they are reprinted in full, distributed at no profit, and include credit to the author, the publication, and the Bank. Library of Congress Catalog Card Number: 77-646559