long-term bonds address the

Peter Fortune

Professor of Economics at Tufts Uni-versity, and Visiting Scholar at theFederal Reserve Bank of Boston. Theauthor is grateful to Gilbert Metcalf,Daphne Kenyon, Richard Kopcke, JoePeek and James Poterba for construc-tive comments.

T his article assesses recent changes in the structure of the munic-ipal bond market. It reviews the tax legislation, judicial interpre-tations, and other factors that affect the yield on municipal bonds.

These factors are then employed in a statistical analysis of the determi-nants of municipal bond yields. A companion piece will examine thepublic policy issues surrounding tax exemption and assess reforms ofthe market. The two papers represent a continuation of research at theFederal Reserve Bank of Boston on the subject (Huefner 1971; Fortune1973; Peek and Wilcox 1986).

The first section of this article is an overview of the key features ofmunicipal bonds, of the most significant changes in the structure of themarket in recent years, and of the question of the constitutional andlegislative basis for tax exemption. The second section focuses on thesignificant features of the income tax code that affect the municipal bondmarket. The third section reviews recent tax legislation affecting themarket. The fourth section presents an econometric analysis of munic-ipal bond yields, designed to determine whether the factors discussed inprevious sections do, in fact, influence yields. The final section is asummary.

I. An Overview of the Municipal Bond Market

Several innovations in the nature of municipal debt have occurredin recent years. The purpose of this section is to outline the basic formsof municipal debt and to describe their yields, their exemption from tax,and their ownership. Petersen (1991) provides an excellent source forinnovations in municipal financial instruments, such as variable-ratebonds and municipal commercial paper.

Forms of Municipal Debt

Municipalities issue debt in a variety of forms.State constitutions and statutes typically restrictshort-term debt to purposes related to working capi-tal, bridging the gap between expenditures and re-ceipts. The most prominent forms of short-term debt,or notes,~ are tax anticipation notes, revenue antici-pation notes, grant anticipation notes and bond an-ticipation notes. Tax, revenue, and grant anticipationnotes are used to provide funds for operating ex-penses, such as payments for wages, salaries, utilitiesand materials. They are repaid from anticipated taxrevenues, federal or state grants, and non-grant,

Long-term bonds address theproblems of the lumpiness of

municipal capital spending andthe intergenerational nature of

benefits.

non-tax revenues, respectively. Bond anticipationnotes are used to provide temporary financing ofcapital outlays, such as purchase of equipment andconstruction of schools and roads. They are repaidfrom the permanent long-term bond financing.

Long-term bonds are issued for permanent fi-nancing of capital outlays, such as construction ofbridges and roads, water and sewage systems, andschools. The purpose of long-term debt is to smoothout the path of tax revenues required to financecapital outlays and to distribute those revenues overtime in conformity with the stream of benefits result-ing from the project. For example, a solid wastedisposal system is "too expensive" to be financed outof tax revenues in a single year, and the benefits ofthe disposal system occur over a long period. There-fore, financing from tax revenues would place a highburden on the current generation of taxpayers but nofinancial burden on future beneficiaries. Long-termbonds provide a way of addressing these problems ofthe lumpiness of capital spending and of the inter-generational nature of benefits.2

The two broad classes of long-term municipaldebt differ according to the source of debt servicepayments (coupons plus principal). General obliga-

tions are backed by the "full faith and credit" of thecommunity, meaning that debt service is to be paidfrom general tax revenues. General obligations are,other things equal, a safe form of investment forindividuals and financial institutions, particularlywhen no limits exist on the ability of the issuer toraise the money via taxes to meet debt service re-quirements.3 Only a few defaults of general obliga-tions have occurred in this century, the most promi-nent being New York City in 1975 (on $2.4 billion ofnotes), Cleveland in 1978 (on $15.5 million of notes),and Bridgeport, Connecticut in 1991.

Revenue bonds have more limited backing--therevenues from specific projects. These bonds areissued by governmental agencies set up to finance,construct, and manage specific facilities. Examples ofthe hundreds of revenue authorities around theUnited States are the Massachusetts Turnpike Au-thority, the Massachusetts Port Authority, the Mas-sachusetts Water Resource Authority and---for a na-tional flavor--the Washington Public Power SupplySystem (WPPSS). If the revenues from the project(auto tolls, airplane landing fees, water and electricitybillings, and the like) are not sufficient to meet debtservice payments, the bonds can be defaulted and theissue goes to the courts to decide which claimants~employees, suppliers, or bondholders--will get paid.The bulk of municipal bond defaults have beenrevenue bonds, the most prominent in recent historybeing the 1983 default by WPPSS on $2.25 billion ofbonds issued to finance nuclear generafing facilities.4

Revenue bonds are the most rapidly growingform of bond indebtedness for states and local gov-ernments. This is due, in part, to restrictions in stateconstitutions that limit the ability of municipalitiesand states to issue general obligation bonds, leaving

1 The term "notes" is usually applied when the maturity of theinstrument is 12 months or less. "Bonds" refer to instruments withmore than one year to maturity.

2 This assumes that full capitalization of tax liabilities intoproperty values does not occur. If full capitalization does exist, inwhich case the tax liability associated with borrowing by a state orlocal government is fully reflected in the value of residential andcommercial property, the residents at the time of the bond issuewill pay the full costs of the debt service regardless of maturity.

~The most default-free securities are, of course, U. S. Treasurybonds, for the government can always print the money to meetdebt service payments if tax revenues are insufficient.

4 One innovation in the past 20 years has been municipal bondinsurance. About 10 percent of the WPPSS default was insured bya private company. The investors in the unit trusts that boughtinsurance did not lose principal or coupons as a result of thedefault.

14 September/October 1991 New Englm, d Economic Review

the alternative of forming revenue authorities. Whilerevenue bonds carry higher interest rates than gen-eral obligation bonds, the use of revenue bonds is theresult of a mutually beneficial arrangement betweenissuers and investors: issuers can finance projectswi~h revenue bonds without ransoming their taxingauthority, paying the higher interest rates with feesand user charges borne by the beneficiaries of theprojects, while investors can get higher rates ofreturn in the form of a risk premium on municipalbonds.

Recent years have seen a rapid expansion ofmunicipal bonds issued for purposes other than thetraditional financing of infrastructure constructed bystates and local governments. Among these are "pri-vate-activity" municipal bonds, advanced refund-ings, and arbitrage bonds. These will be discussedbelow.

Eligibility for Tax Exe~nption

The right to issue tax-exempt bonds is not auto-matic. In order to be eligible for exemption, a bondmust meet standards imposed by law in section 103 ofthe Internal Revenue Code, as interpreted by theInternal Revenue Service. The purpose is, of course,to prevent municipalities from using tax-exempt debtto finance projects judged not to be for municipalpurposes.

The problem of "private-activity" municipalbonds has grown dramatically in the last two decades(Zimmerman 1991). In the 1960s municipalities beganto use their favored access to credit markets to inducebusinesses to locate within their jurisdiction. Thedevice employed was the industrial developmentbond, issued by a municipality to finance construc-tion of structures that were then leased to the privatebusiness, the lease payments providing the funds tomeet debt service payments. In this way a businesscould finance its construction of factories and officespace at municipal interest rates rather than at corpo-rate bond rates.

The success of tax-exempt industrial develop-ment bonds led to a plethora of similar revenuebonds, each designed to serve a specific constituency:for example, mortgage revenue bonds, to financeloans at below-market rates to households to pur-chase homes; student loan bonds, to make loans tostudents at favorable rates; and pollution controlbonds, to provide low-cost funds to corporations toacquire equipment for the reduction of water and airpollution. By the early 1980s the issuance of private-

activity bonds was out of control. This can be seen inFigure 1, which shows the share of all long-term debtoutstanding (corporate and foreign bonds, municipalbonds and mortgages) that is tax-exempt, whether forpublic or private purposes. The figure shows thattotal tax-exempt bonds declined as a share of allbonds and mortgages from 1960 through 1990, with abrief surge in the early 1980s. The state-local govern-ment share for public purposes declined throughoutthe period, while the private-activity share (issued onbehalf of households and corporations) rose begin-ning in the early 1970s, growing to almost one-thirdof outstanding tax-exempt bonds by 1985.

The rapid growth of private-activity tax-exemptbonds led to the realization that those issues werehaving effects not intended by Congress. First, thefederal taxpayer was underwriting a capital-cost sub-sidy for households and corporations, a subsidynever intended by Congress. Second, the competi-tion for tax-exempt credit from private-activity bondswas forcing interest rates up for the intended benefi-ciaries of tax exemption--issuers of public purposebonds.

A significant feature of the Tax Reform Act of1986 was legislation limiting the eligibility of private-activity bonds for tax exemption. While limits hadexisted in earlier years, they were tightened consid-erably by this Act. Section 103(c), under which theprivate-activity definition is established, applies four

A significant feature of the TaxReform Act of 1986 limited the

eligibility of private-activity bondsfor tax exemption.

different tests to determine whether a bond is private-activity. The first test--which is the most widelyapplied--is a joint test of uses-of-funds and of secur-ity interest: if more than 10 percent of a bond’sproceeds are used for private purposes and if morethan 10 percent of the debt service is derived, directlyor indirectly, from a private use, the bond is deemeda "private-activity" bond.

In order to be tax-exempt, a private-activity bondmust (with some exceptions) be within the statevolume limit and it must satisfy certain maturity

September/October 1991 New England Economic Review 15

restrictions. The-state volume limit, established in1986, is now set at the greater of $50 per capita or $150million. Tax exemption is automatically extended toprivate-activity bonds within that limit. However,seven categories of private-activity bonds are ex-cluded from the volume limit and can be issued intax-exempt form in any amount. These are "exemptfacility bonds" (airports, docks and wharves, solidwaste facilities, and the like), qualified mortgagebonds, qualified veterans’ mortgage bonds, qualifiedsmall-issue industrial development bonds, qualifiedstudent loan bonds, qualified redevelopment bonds,and qualified section 501(c)(3) nonprofit organizationbonds.

Figure 1 shows a dramatic reversal of the earliertrends after the 1986 Tax Reform Act. Following abrief surge in private-activity issues in 1985 in antic-ipation of the limitations imposed in the 1986 Act, theoutstanding amount of tax-exempt bonds for privatepurposes declined sharply.

Another problem category of bonds is arbitragebonds. In the 1960s, states and local governmentsbecame aware that tax exemption was a moneymachine: by issuing tax-exempt bonds at favoredinterest rates and investing the proceeds in taxablebonds,, a municipality could earn a spread equal tothe difference between the taxable and tax-exemptrates. Note that while these arbitrage profits can aidthe states and local governments by providing fundsto finance additional public services or to reduce taxrates, they are not necessarily in the financial inter-ests of the taxpayers: that depends on the differencebetween the municipal bond rate paid and the after-tax rate of return the taxpayers earn on their invest-ments.

In 1969, section 103(c) of the Internal RevenueCode was changed to define arbitrage bonds as bondswhose proceeds, beyond a "reasonably required"reserve, were used to invest in securities with a"materially higher yield" for more than a "temporaryperiod." The problem of arbitrage bonds is presentedmost clearly when a bond is issued with the solepurpose of investing the proceeds at a higher interestrate and earning the spread over the period until thebond is repaid; this blatant use of tax exemption wasclearly eliminated by the 1969 legislation. However,over the intervening 20-odd years the interpretationof section 103(c) has changed frequently. This is theresult of two conflicting goals, the first being to closeoff the arbitrage opportunities still employed despitethe regulation, and the second to allow tax exemptionfor bonds issued for "reasonable" uses, when the

Figure I

Tax-Exempt Share of OutstandingBonds and Mortgages

20 Pe~cenl

Total Tax-Exempl Share

Govemrnent Share

18

16

14

12

10

8

6

4

2~

1960 1964 1968 1972 1976 1980 1984 1988

Source: Board of Governors of theFederal Reserve System, Flow of Funds Accounts

proceeds are temporarily invested in earning assets,sBecause municipalities often find themselves in thesituation of issuing a bond before the proceeds areexpended, temporarily holding the proceeds, thedifficulty in the arbitrage restrictions is setting stan-dards that allow this to happen for reasonable pur-

s An example is advanced refunding: the issuance of a newmunicipal bond to pay off an outstanding bond. The proceeds ofthe new issue are typically placed in an escrow account that holdsU.S. Treasury securities whose income is sufficient to pay the debtservice on the advance refunding bonds and to yield net income.The amount of the proceeds of the advance refunding issue, plusaccumulated net income, is designed to be sufficient to pay off theoriginal bond. Clearly this must be done before the retirement ofthe original issue at either maturity or a call date.

The optimal timing for an advanced refunding is when interestrates are low, so the advance refunding is often done well beforethe actual retirement of the outstanding issue. If the delay issufficient to arouse the interest of the Internal Revenue Service, thenew issue can be interpreted as an arbitrage bond and, in order tomaintain the tax-exempt status of the advance refunding issue, themunicipality must rebate to the U.S. Treasury the spread betweenthe interest earned on the Treasury securities and the interest paidon the newly issued municipal bond. If this rebate is not made, theinvestor will lose the tax exemption, even if he had every reason tobelieve that the exemption was allowed when he purchased thebond.

16 September/October 1991 Nezo England Economic Review

poses, while eliminating the abuses of tax exemp-tion.6

The Question of Constitutionalit~dThe exemption of state and local interest pay-

ments on municipal bonds from federal income tax-ation has a long history, originating in a questionabout the legality of taxation of activities of one levelof government by another level of government. Dur-ing the years immediately following the AmericanRevolution, power rested in the states and the federalgovernment was weak. Concern for the financialfragility of the new federal government, combinedwith several attempts by states to tax activities of thefederal government, led to a series of Supreme Courtdecisions, under Chief Justice Marshall, that pro-tected the central government from the taxing powersof the states. Marshall’s well-known dictum that "thepower to tax is the power to destroy" was, in fact, astatement designed to protect the central govern-ment, although in modern times it has been used tosupport protection of the states from federal taxingpowers. One of the most important of these caseswas McCulloch v. Maryland (1819), in which Marshall’scourt struck down a tax levied by the state of Mary-land on the Bank of the United States. This was theorigin of the exemption of federal activities from statetaxation.

The first federal income tax was enacted duringthe Civil War. It taxed both salaries and interestpayments by states and local governments, but inCollector v. Day (1871) the Supreme Court ruled thatthe application of the tax to the salary of a state judgewas unconstitutional. This decision established thedoctrine of reciprocal immunity, in which the federalgovernment and state-local governments were pro-tected from the tax powers of the other. The expira-tion of the federal income tax in 1872 meant that thisdecision had a limited effect, but enactment of a newfederal income tax in 1894 revived the issue. The 1894federal income tax explicitly recognized Collector v.Day by exempting salaries paid by state and localgovernments, but included interest payments bystates and local governments in the tax base. Thefederal taxation of state and local interest paymentswas struck down in 1895 in Pollock v. Farmers’ Loan &Trust Company, when the U.S. Supreme Court heldthat interest on a state bond should be exemptedfrom federal taxation. This case has been central tothe argument that the exemption is protected by theConstitution.

The modern federal income tax was ushered inwith the Sixteenth Amendment to the Constitution in1913. The Sixteenth Amendment gave Congress thepower "to lay and collect taxes on income, fromwhatever source derived." In the new income tax,enacted after the Sixteenth Amendment, the questionof reciprocal immunity was avoided by explicitlyexempting both interest and salaries paid by statesand local governments, thus adhering to both Collec-tor v. Day and Pollock v. Far~ners’.

Because the initial tax rates were very low, theeffect of this exemption on the state-local cost ofcapital was small. However, as federal income taxrates rose, the exemption of municipal interest be-came an economic issue. The 1920s and 1930s wereperiods of considerable debate about the interestexemption, with parties lined up on the basis ofeconomic self-interest. Business organizations, suchas the U.S. Chamber of Commerce, private utilitycompanies that opposed a subsidy to their publiccompetitors, and many state governments in theindustrial North joined to oppose tax exemption. Onthe other side, many state governments in the non-industrial South strongly supported the exemptionbecause of the subsidy it conferred and because theyhad little industrial base to be harmed by the compe-tition for funds.

A constitutional amendment to eliminate theexemption failed in 1922, but it continued to be urgedby the Coolidge, Harding, and Hoover Administra-tions. Gradually the scope of the exemption wasnarrowed by judicial decision rather than legislation.In Graves v. New York ex tel. O’Keefe (1939) the Courtupheld a state tax on a federal salary, thereby over-turning Collector v. Day. Ultimately this paved theway to elimination of the intergovernmental exemp-tion for salaries. However, the exemption continuedfor interest payments, primarily as a result of theefforts of state and local governments.7

Proponents of the view that the exemption isprotected by the Constitution have appealed to earlySupreme Court decisions, such as Pollock v. Farmers’Loan & Trust Company, and to the proposals in the

6 For much of the life of the arbitrage regulations, a munici-pality could avoid the arbitrage bond limitations by investing nomore than 15 percent of the bond proceeds in a reserve fund, andby showing "due diligence" in completing the project the bondwas intended to finance, with an upper limit of five years on thetemporary holding period (Buschman and Winterer 1983). This stillallowed considerable leeway for earning arbitrage profits, and apart of the 1986 Tax Reform Act tightened the restrictions further.

7 This history of the early debate is drawn from Huefner(1971).


1920s to eliminate the exemption through a constitu-tional amendment. A significant minority of the legalprofession continues to assert that taxation of state-local interest is barred by the U.S. Constitution, eventhough that uncertainty was eliminated in a series ofU.S. Supreme Court decisions in the 1980s. Whateverone’s views of the constitutionality issue, however,legislation has clearly conferred tax exemption: sec-tion 103(a) of the Internal Revenue Code of 1954, thebasis for the present tax code, specifically exemptsmunicipal interest income from the federal incometax.

Whatever one’s view of theconstitutionality issue, legislationhas clearly conferred tax exemp-tion for state and local interestpayments on municipal bonds.

tutional foundation for it. In Garcia v. San AntonioMetropolitan Transit Authority (1985), the Court heldthat the taxation of state and local interest paymentswas a matter of legislation, not a constitutional issue:if Congress wished to tax municipal interest, it wasfree to do so. This was upheld and clarified in a 1988Supreme Court decision, South Carolina v. Baker,which upheld the 1982 TEFRA removal of tax exemp-tion for state and local bonds not held in registeredform. In his majority opinion, Justice Brennan arguedthat "states must find their protection from Congres-sional regulation through the national political pro-cess, not through judicially defined spheres bf unreg-ulated state activities."

Thus, at this point no constitutional barrier toelimination of the exemption of municipal interestremains. However, Congress has given no indicationof interest in including municipal interest in thefederal income tax base. The reason is that tax exemp-tion has become a very valuable subsidy for statesand local governments, and states and local govern-ments have formed a powerful lobby to protect theirpreferred access to credit markets.

This legislative basis for tax exemption has beenweakened in recent years. The first important deci-sion addressed the Social Security Act Amendmentsof 1983, which imposed a tax on Social Securitybenefits if the taxpayer’s income was above a speci-fied level. For the purposes of this computation,"income" was defined as including interest frommunicipal bonds. As a result, the amendments indi-rectly imposed a tax on municipal interest: sufficientmunicipal interest income could lead to payment ofadditional taxes. The Supreme Court refused to hearcases charging that this was unconstitutional, therebyallowing continuation of the indirect taxation of mu-nicipal interest.

The second important piece of legislation weak-ening tax exemption was the 1982 Tax Equity andFiscal Responsibility Act (TEFRA), which limited theexemption to municipal bonds issued in registeredform, forcing all bearer bonds into the taxable cate-gory.8 The purpose of this was to allow the federalgovernment to track municipal interest income, anobjective furthered in 1987 when the federal incometax forms required the taxpayer to report his munic-ipal interest income even though it was not subject totax.

Just when Congress was weakening the exemp-tion, the Supreme Court clearly rejected any consti-

Unusual Features of Municipal Bonds

Municipal bonds differ from corporate, U.S.Treasury, and other taxable bonds in ways other thantax exemption. Perhaps the most prominent is theserial form of municipal bond issues, which affectsboth the liquidity and trading costs of municipalbonds.

Corporate and Treasury bonds are typically orig-inated as a single issue in which each certificatecarries the same coupon rate, call date, and maturitydate. While retirement schedules typically differ--some corporate bonds are retired in a lump sum,while others are retired gradually through sinkingfunds--the essential feature of a taxable security is asingle issue broken up into a large number of certif-icates, each with the same characteristics.

Municipal bonds, on the other hand, are typi-cally sold in a "serial" form in which each "strip" is aseparate bond. This means that instead of a singlelarge issue, the bond consists of a series of smallerissues, each with its own maturity. For example,suppose that a city wants to borrow $100 million to

8A bond is held in registered form if the owner’s name isrecorded by a transfer agent--usually a bank-~whose function it isto keep ownership records. The other form is a bearer bond,deemed to be owned by whoever holds it and for which no recordof ownership is kept.


construct a sewage system. Instead of issuing one20-year bond, which trades as a unit, it might issue 20bonds, each having a different maturity: the first"strip" has a one-year maturity, the second strip hastwo years to maturity, and so on until the twentiethstrip, which has 20 years to maturity. The distributionof the total amount between strips is a financialdecision that the city must make at the time of theissue; for example, the city might want to retire anequal amount in each year, so each strip would havea par value of $5 million.

The serial bond is sold as a block to the winningunderwriting syndicate. But when the syndicate sellsthe issue, it splits the issue into the individual strips,and the individual strips into certificates, which aremarketed separately. The advantage of this approachis that the underwriter can typically get a higher pricefor the entire issue by selling each strip to investorswho most want that maturity, and the issuer cantailor the retirement schedule to its needs by decidinghow much of the issue will be allocated to each strip.If each strip is small, however, the cost of secondarymarket trading in individual strips might be higherthan if the issue were traded as a unit, and this highertransaction cost will be reflected in a lower price paidby the investor.

Figure 2

Interest Rate Ratios for SelectedTerms, Prime Municipal Bonds

vs. U.S. Treasury Bonds

Interest Rate Ratio

t

.4

1965 1969 1973 1977 1981 ~985 1989

Source: Salornon Brothers, Inc.

Yields on Tax-Exempt Bonds

Exemption from federal income taxes confers anadvantage upon the holder of municipal bonds. Thatadvantage leads investors to require lower before-taxyields on municipal bonds than on taxable bonds ofequivalent maturity and risk. The result is that theyield to maturity on tax-exempt bonds is less than theyield on taxable bonds with equivalent maturity andquality.

This advantage is shown in Figure 2, whichreports the ratio of the yield on prime-grade munici-pal bonds to the yield on UoS. Treasury bonds. This"interest rate ratio" is shown for one-year, five-yearand 20-year bonds.9

Several important observations can be drawnfrom Figure 2. First, for each of the three maturitiesshown, the interest rate ratio is less than unity,reflecting the tax advantages of municipal bonds.Second, for each maturity the interest rate ratio ishighly variable: at times the advantage to the issuer oftax exemption appears to be quite small, as in 1969when 20-year municipal bonds carried yields almostequal to 20-year Treasury bonds, while at other timesthe advantage is very great, as in the late 1970s when

the interest rate ratio for 20-year bondswas about0.65.1°

Finally, the yield curve for municipal interestrates rises more rapidly than the yield curve forTreasury securities. This is evident from the fact thatthe interest rate ratio is higher for five-year bondsthan for one-year bonds, and higher for 20 years thanfor five years. This phenomenon almost disappeared

9 Prime-grade municipal and U.S. Treasury bonds are notprecisely equivalent: Treasuries are of higher credit quality becausethe federal government can print the money necessary to repay itsdebt; Treasuries trade at lower transactions costs because theyhave a single issuer, while municipals are issued by a variety ofstates, local governments, and authorities whose quality is difficultto determine and whose issue sizes can be small; Treasuries aretraded in a thick market while municipals have a lower level ofliquidity. But these differences are (arguably) sufficiently small toallow us to use the interest rate ratio as a measure of the influenceof tax exemption on municipal bond yields.

10 The 1969 experience is an anomaly, probably due to thedebate surrounding the 1969 Tax Reform Act. The House proposalswould have included state-local interest in the new AlternativeMinimum Tax. While the AMT was adopted, state-local interestincome was not subject to the AMT. However, the prospect thatthese proposals would be adopted led to very high interest rateratios in that year. For a discussion of the debate, see Huefner(1971), p. 100 ft.


in the late 1980s, especially for short- to intermediate-term bonds. Both the more rapidly rising yield curve(relative to Treasuries) of 1966-85, and the elimina-tion of the differential for short- to intermediate-termbonds, are due to the role of commercial banks in themunicipal bond market. These changes--whichmean that the municipal bonds appear to trade morelike taxable bonds--have been particularly markedsince the Tax Reform Act of 1986.

Changes in Municipal Bond Ozonership

Prior to the mid-1980s the largest holders oftax-exempt debt were financial institutions, primarilycommercial banks and property and casualty insur-ance companies (Figure 3).11 Indeed, from 1962through 1980 the share held by households declinedfrom over 40 percent to about 25 percent, with mostof the market share moving toward commercialbanks. State-local governments (primarily retirementfunds) were significant holders of municipal bonds inthe early 1960s, with about 10 percent of the out-standing stock, but this share declined thereafter.Insurance companies (primarily property and casu-alty insurers) have maintained a 15 to 20 percentshare of outstanding state-local bonds, with a briefsurge in 1975-80 and a return to their normal share by1985.

Beginning in the mid-1980s the ownership pat-tern of tax-exempt bonds changed dramatically, with

Beginning in the mid 1980s theownership pattern of tax-exemptbonds changed dramatically, with

commercial banks sharplyreducing their share and the share

held by households increasing.

commercial banks sharply reducing their share fromalmost 55 percent in 1980 to about 25 percent by 1990.The withdrawal of commercial banks from the marketwas matched by an increase in the share held byhouseholds, from about 25 percent in 1980 to over 60percent in 1990. The nature of household ownershipalso changed, as unit investment trusts and mutualfunds provided both diversification opportunities

and liquidity services not available to most house-holds prior to the 1980s.

The tax structure provides some explanation ofthese changes in ownership. Until the 1982 EconomicRecovery Tax Act (ERTA), households faced taxrates as high as 70 percent, while commercial banksand property and casualty insurance companies faceda tax rate of 46 to 52 percent. Furthermore, until1982 commercial banks could deduct from their tax-able income all interest paid to carry municipal secu-rities, a tax advantage that gave them a strong incen-tive to hold tax-exempt debt. Thus, higl~-incomehouseholds, commercial banks, and property andcasualty insurance companies had the greatest incen-tives to hold tax-exempt bonds and were the domi-nant holders. 12

The 1982 Tax Equity and Fiscal Responsibility Actreduced the deductibility of interest expense forbanks carrying tax-exempt securities to 85 percent ofthe interest paid. The 1984 Deficit Reduction Actfurther reduced the tax advantage for banks to 80percent of carrying costs. Finally, with some excep-tions,13 the 1986 Tax Reform Act completely elimi-nated any deduction by banks of interest paid to carrymunicipal bonds, virtually extinguishing the advan-tages for banks of tax-exempt securities. Figure 3shows the banks’ share of the market stabilizing inthe early 1980s and falling sharply after 1985. Theplunge after 1985 is the result of the sharp decline incorporate income tax rates and the elimination ofbanks’ deduction of interest paid to carry municipalbonds.

The Effect of Future Tax Rates on Bond Yields

The bond yields that prevail at any moment willbe affected by anticipations of future tax rates: ifinvestors believe that tax rates will rise (fall) in thefuture, they will require a lower (higher) yield onnewly issued municipal bonds, relative to the yield

11 Property and casualty insurance companies were by far themost important insurance companies in the municipal bond mar-ket. Until recently, life insurance companies faced relatively lowtax rates, limiting their interest in municipal bonds.

12 The property and casualty insurance sector’s share of themarket remained low in spite of high statutory tax rates becausethose companies experience a strong cycle in profitability. Thus,the tax advantages of municipals were enjoyed only during prof-itable years. This reduced the effective tax advantages.

13 Municipal bonds acquired before August 16, 1986 still haddeductible carrying costs. This led to a surge in bank purchases ofprivate purpose bonds in anticipation of grandfathering of theexemption. Also, deductibility was continued for bank purchasesof small local issues.

20 September/October 1991 New England Economic Review

Figure 3

Shares of Outstanding State and LocalBonds Held by Selected Sectors

Commercial Banks50 ~ ~

State-Local Government~0, , , ", , ,-m~, ,-~, _~---;- ,-~.-----,1960 1964 1968 1972 1976 1980 1984 1988

Source: Board of Governors of the Federal Reserve System,Flow of Funds Accounts, quar[erly data.

on newly issued taxable bonds. It is difficult toquantify the role played by anticipations of tax rates:nobody keeps a record of what "the market" thinkstax rates will be in the future. One approach is to use"event analysis" to infer the influence of anticipatedtaxes. Poterba (1986) and Fortune (1988) have inves-tigated the behavior of interest rates at times of taxpolicy changes, and find that the behavior of interestrates is consistent with the influence of anticipatedtax rates on municipal bond yields.

Another approach is to assume that the marketcorrectly anticipates the future and use these esti-mates as "actual" future tax rates to infer the effect oftax rate anticipations at any moment. However, thenotion that--on average--the market is correct hasreceived considerable attention, and very little sup-port, from academics. For example, Fortune (1991a)surveys the evidence from the stock market, conclud-ing that "the efficient market" is a concept worthselling short.

A third approach, adopted here, is to use marketdata to infer tax rate anticipations. One source for thisis the yield curve on municipal and taxable securities.The implied marginal future tax rate for municipalbond investors can be derived as follows. Assume

that an estimate is sought for the tax rate anticipatedby the marginal investor in municipal bonds betweenfive and ten years from now. If the interest rates areknown for five-year municipal and taxable bonds thatare newly issued five years from now, it is possible toestimate the implied tax rate for five to ten years inthe future. Denoting RM,5,1o as the yield on a munic-ipal bond bought five years from now that maturesten years from now, and RT, s,10 as the yield on anequivalent taxable bond, the implied future tax rate ists4o = 1 - (RM,5,10/RT,5,10).

Unfortunately, no direct information on futureinterest rates is available. The futures market inmunicipal bonds does not extend very far into thefuture and it is based on a bond yield index, not onyields for specific securities. However, indirect infor-mation can be found in the term structure of interestrates. According to the Expectations Hypothesis, themost widely held theory of the term structure, theyield to maturity on a ten-year municipal bondbought now (RMo0A0) is mathematically related to theyield on a five-year bond bought now (RM,0,5) and theexpected yield on a five-year bond bought in fiveyears (RM,5,10). The equilibrium relationship is:

(1) (1 + RM,0,10)1° = (1 + RM,0,5)5(1 + Rlvl,5,10)5.

According to this relationship, $1 invested nowin a ten-year bond will have an accumulated value inten years equal to the accumulated value in ten yearsof a sequence of two investments: $1 invested now ina five-year bond, followed by investment of thataccumulated value in a second five-year bond matur-ing five years from that date of purchase. Equation (1)assumes that the bonds are zero-coupon bonds.

The bond yields that prevail atany moment will be affected by

anticipations of future tax rates.

If the relationship did not hold, investors wouldnot diversify their portfolios by holding both five-year and ten-year bonds. Consider investors with aten-year horizon. If the left side of equation (1)exceeds the right side, those investors would holdonly ten-year bonds because the terminal valuewould exceed the terminal value from buying five-


year bonds and reinvesting the proceeds at maturityin five-year bonds. If, on the other hand, the left sideis less than the right side, the investors would buyfive-year bonds and reinvest in five-year bonds atmaturity. Only when the equality in equation (1)holds will investors hold both five-year and ten-yearbonds.

This relationship allows us to derive an estimateof the anticipated yield on five-year municipal bondsin five years. Solving equation (1) for RM,5,1o gives:

(2) RM,5A0 = [(1 + RM,O,m)2/(1 + RM,0,5)] - 1.

Because the values of RMoO, IO and RM,O,5 are

known right now, equation (2) can be used to calcu-late the value of RM,5A0 which is implicit in the yieldcurve. This can also be done for taxable securities. Anestimate of the tax rate expected to prevail betweenfive and ten years in the future can then be obtained:

(3) t5-10 = 1 - (RM,5,10/RT,5,10).

This approach is subject to a number of criti-cisms. First, it assumes that the taxable and tax-exempt yields used in the calculations are for securi-ties that are equivalent in all respects except taxexemption. This is unlikely, because the issuers oftaxable and tax-exempt bonds are inherently differentin terms of default risk, because call features or otherspecial features might make the bonds different, andbecause transactions costs can differ, with municipalbonds generally less liquid and traded in a thinnermarket than taxable bonds.

Second, this approach assumes that investors areindifferent to the maturity structure of their portfolio;no "market segmentation" exists. All investors careabout is the expected yield on their portfolios, andmarket risks are irrelevant to portfolio decisions.

Using the Salomon Brothers data for five-yearand ten-year prime municipal bonds and U.S. Trea-sury bonds, the estimates of t~_10 have been derived(Figure 4). The implied future tax rate shows a goodbit of noise in it, as one would expect because of thepotential for factors such as call features to distort theyield curve relationships. But the general outlines ofthe series seem to fit reasonably well: the impliedfuture tax rate fell sharply after 1979, in apparentanticipation of the tax rate reductions in the 1981Economic Recovery Tax Act (ERTA). Furthermore,after 1982 the implied future tax rate is about 15 to 20percent, consistent with the low tax rates introducedin the 1981 ERTA and subsequent tax acts. Also, the

Figure 4

Future Tax Rate, Five to Ten Years Out

Pe~cen[45

40

35

30

25

20

15

10

5

01970

III1II/" ill ~/tll ill/rli’!1 II I"l ’"llrll’ "

" III1973 t976 1979 1982 1985 1988 1990

Source: Salomon Brothers, Inc. and author’s calculations.

mid-to-late 1970s show a high anticipated future taxrate, a result not inconsistent with the discussions ofthe times, an example being President Carter’s state-ment that "the tax system is a disgrace to the humanrace." It is no surprise that this period of highanticipated tax rates was also a period of low interestrate ratios.

II. The Income Tax Code and MunicipalBonds

The previous section touched on some aspects ofrecent tax legislation to explain major changes in themunicipal bond market. This section reviews the keyfeatures of the income tax code that influence thedemand for tax-exempt debt, and, therefore, help toexplain municipal bond yields. This journey throughthe income tax code provides background essential tothe formal analysis of the determination of municipalbond yields in later sections.

Taxation of Ordinary Income from Bonds

The Internal Revenue Code identifies two forms


of taxable income. The first is "ordinary income,"which consists of cash payments such as coupons,interest paid, cash divide,n, ds, and origi,n, al issue dis-counts,l~The second is ’capital gains, defined asthe difference between the price at which an asset issold and its "cost basis" at the time of sale. The codeallows the holder of a tax-exempt bond to excludeordinary income but not capital gains from taxableincome.

Original issue discount is the difference betweenthe par value of the bond--the amount paid atmaturity--and the initial price at which the bondcame to market. For example, suppose a bond has apar value of $1000 and was originally issued at a priceof $900. The difference of $100 occurs because thecoupon rate on the bond is less than the interest rateprevailing on similar securities at the time of issue.The tax code treats this as cash income even though itdoes not give rise to cash payment until the bondmatures. If the bond is taxable, the discount is in-cluded in taxable income as it accrues; if the bond istax-exempt, it is not included in taxable income.

The tax code does not allow the investor to waituntil the bond matures before he "earns" the originalissue discount. Instead, investors must amortize thediscount over the life of the bond. If the bond’sincome is taxable, the amortized value is added toeach year’s taxable income; if the bond is tax-exempt,it is not added to taxable income. In either case, theamortized value of the original issue discount isadded to the cost basis of the bond and becomesrelevant to the calculation of capital gains.15

As noted above, the amortization of the originalissue discount also affects the cost basis of the bond.The amount of the discount assigned to each year isadded to the cost basis of the bond, so that theaccumulated discount is not treated as capital gainswhen the bond matures or is sold. The logic isstraightforward: if the discount was earned in thepast, it was either taxed if the bond was taxable ortreated as tax-exempt if the bond was tax-exempt. Butif the cost basis is not adjusted upward to reflect theaccumulated original issue discount, when the bondis sold that amount will be treated as capital gains~and capital gains are taxed whether the bond istax-exempt or not!

Taxation of Capital Gains

The Internal Revenue Code taxes capital gains ona cash basis rather than on an accrual basis. Thismeans that gains are taxed only when realized upon

the sale of the asset, not as they accrue "on paper"over the holding period. As noted above, the amountof capital gains is defined as the difference betweenthe price at which the asset was sold and the "costbasis" of the asset. The cost basis is usually the price(including brokerage commissions) at which theasset was purchased, but the cost basis can alsoreflect certain adjustments such as the original issuediscount.

The effect of the opportunity to defer capitalgains taxes is to reduce the effective tax rateonaccrued gains. For example, suppose that during 1991a stock in your portfolio increases by $10 per share. Ifcapital gains were taxed on an accrual basis, youwould have to include that $10 in your 1991 taxableincome even though you had not sold the asset. Butsuppose that you do not sell the stock until 1996.Assuming a capital gains tax rate of 31 percent, youwill pay a tax of $3.10 per share in 1996, but at aninterest rate of 10 percent, the present value in 1991 ofyour 1996 capital gains tax liability is only $1.92.Thus, your effective capital gains tax rate in the yearof accrual is only 19.2 percent if you defer the gainsfor five years.16

Until recently, capital gains were taxed at a lowerrate than ordinary income. For example, from 1942 to1978, the tax code allowed an investor to exclude 50percent of capital gains from taxable income, thussetting the capital gains rate at 50 percent of theordinary income tax rate. During much of that time,however, the maximum capital gains tax rate was setat 25 percent. From 1978 to 1986 the code allowed 60

14 Coupons, interest, and cash dividends are, with someexceptions, defined as ordinary income and included in a taxpay-er’s taxable income in the year they are received unless they areexplicitly excluded, as in the case of a tax-exempt bond. Among theother exceptions are return of capital dividends and a majorportion of dividends received by financial institutions.

15 After the 1982 Tax Equity and Fiscal Responsibility Act(TEFRA), original issue discount was amortized as accrued inter-est: the investor calculated the interest rate that would accumulatethe initial price to be equal to the price at maturity, and an amountequal to that interest rate times the cost basis was treated asordinary income in each year. Before TEFRA the amortization wasconstant in each year.

16 Indeed, capital gains taxes can be avoided completely byholding assets until death, because the cost basis of assets isadjusted to the market price at the time of death, effectivelyeliminating any taxation of the capital gains when heirs sell theassets. However, the potential avoidance of taxation of capitalgains as income does not mean that the gains are shielded from alltaxation: federal estate taxes are levied if the estate is large enough.Because the estate tax liability depends upon the size of the estate,not upon the amount of the capital gains, two estates of equal sizewill pay the same tax even though one might have a considerablygreater appreciation component than the other. Thus, the estatewith the greater capital gains will not pay any additional taxes.

September/October 1991 Nezo England Economic Reviezo 23

percent of gains t6 be excluded, which, because of the50 percent maximum tax rate on ordinary income,had the effect of reducing the maximum tax rate oncapital gains to 20 percent. Thus, an individual tax-payer in the highest bracket would pay 50 percent onordinary income but only 20 percent on capitalgains.17

The Tax Reform Act of 1986 eliminated thisdifferential. Under that law, 100 percent of capitalgains are included in taxable income and, with a 33percent maximum tax rate on ordinary income, the

The effect of the opportunity todefer capital gains taxes until thesale of the asset is to reduce the

effective tax rate on accrued gains.

maximum capital gains tax rate was also 33 percent.The Revenue Reconciliation Act of 1990 restored aslight differential in favor of capital gains by setting amaximum capital gains tax rate of 28 percent.

Note that capital gains taxation makes municipalbonds selling at a discount less attractive because theinvestor is taxed on part of the return. To see this,assume that an investor is choosing between a newlyissued taxable bond and a newly issued tax-exemptbond, each priced at par. He knows that if interestrates rise, the price of each bond will fall, and that ifhe sells the bond at a loss, the capital loss is deduct-ible regardless of whether the bond is taxable ortax-exempt. But the market discount at the time ofsale will expose the new buyer to capital gains taxes.Because part of the return to the municipal bond willnow be taxable, the new buyer will pay a still lowerprice to compensate him for the capital gains taxes.The taxation of capital gains is, however, not disad-vantageous for the taxable bond since the new buyerwill pay taxes whether his income is in the form ofcoupons or capital gains.

Therefore, the prospect of price decreases shouldmake municipal bonds less desirable than taxablebonds, inducing a higher interest rate ratio. Thiseffect should have been greater prior to the 1986 TaxReform Act, when capital gains were taxed at lowerrates than ordinary income. Furthermore, the effectof the taxation of capital gains on municipal bondsshould also be greater when interest rates and asset

prices are more volatile. The econometric analysis inSection IV will employ a measure of price variabilityto determine whether this hypothesis is supported.

Taxation of Capital Losses

The tax treatment of capital losses is not symmet-rical with the treatment of gains. While 100 percent ofrealized gains are included in taxable income, the taxcode might not allow full deduction of losses fromtaxable income if losses exceed capital gains: for amarried couple filing jointly, losses can be fullydeducted up to the amount of capital gains plus$3,000, but any losses above that amount must becarried over to future years. Thus, while a taxpayercan ultimately deduct all losses so long as taxableincome is sufficient to take the deduction, the ceilingon the loss deduction allowed in a single year reducesthe present value of the tax saving by the requirementto defer the deduction to future years.

For example, consider a married person filingjointly who has taxable ordinary income of $20,000.Suppose that this person has a capital gain of $2,000and a capital loss of $10,000. In the current tax year hecan deduct $5,000 of capital losses--S2,000 againstthe capital gain plus the $3,000 maximum loss offset.The remaining $5,000 of capital losses can be carriedover to future years: $3,000 can be deducted in thenext year, and the remaining $2,000 in the followingyear.

The absence of full loss offsets leads investors toalter their portfolios to reduce their exposure to pricerisk. This does not mean that they should avoid riskyassets, only that the absence of full loss offset pro-vides an incentive to reduce investments in riskyassets below the level that would prevail with full lossoffset.

State Income Taxes

The focus of this paper is the federal income taxcode, but no survey of the connection between taxa-tion and the municipal bond market is completewithout reference to state income taxation. Forty-twostates levy an income tax on interest and dividends,

17 Since 1969 an Alternative Minimum Tax has applied inunusual circumstances and can have the effect of increasing the taxrate on capital gains. The Alternative Minimum Tax does not havea substantial effect on the tax positions of most investors, and hasnot been considered in this overview of capital gains taxation.


with tax rates ranging from very low levels to a highof 12 percent in Massachusetts and North Dakota.18In addition, 36 of the states with an income taxexempt some or all interest paid by governmentagencies within their own jurisdiction, and tax allout-of-state municipal interest income. Thus, in moststates in-state municipal bonds offer a tax advantage.

If the marginal investor in a state’s bonds is aresident of the state, jurisdictions in those states willpay lower municipal bond rates than those paid instates with lower tax rates. This will occur becausethe intra-state investors will pay a higher price forthose bonds than will investors in other states. How-ever, if the marginal investor is out-of-state, highstate income tax rates will not affect the municipalbond yields paid by jurisdictions within the state. Inthis case, the advantages of within-state tax-exemptbonds accrue as a "windfall" to within-state inves-tors.

The importance of this segmentation has beenexamined by Kidwell, Koch and Stock (1984), whoconclude that--as a generalization--a state’s resi-dents are the marginal investors in that state’s state-local bonds. Therefore, they find that (other thingsequal) high state income taxes do confer lower bor-rowing costs on the state and its political subdivi-sions.

IlL Recent Tax Legislation

The key events in legislation in recent years aresummarized in the box. The most prominent changeis a radical revision of tax rate schedules, with both ageneral reduction in tax rates and a reduction in thedegree of progressivity. The rate reductions shouldlead to an increase in interest rate ratios, while thedecline in progressivity should reduce the rate ratios.In addition, since 1985 the tax rate schedule has beenindexed, thereby reducing the "bracket creep"which, slowly and over time, pushed tax rates up andreduced the interest rate ratio. Finally, specificchanges have limited the desirability of municipalbonds to financial institutions and restricted the useof tax-exempt bonds for "private" purposes. Thesechanges will be discussed in order.

Changes in Tax Rates

Federal taxation of corporate income establishesa lower tax bracket for corporations with very low netincomes, but most corporate income is taxed at the

maximum statutory tax rate. That rate changed verylittle until 1987: the maximum tax rate on corporateincome was 52 percent in the early 1960s, fell to 50percent in 1964, to 48 percent in 1965, then to 46percent in 1979, then to the present 34 percent in1987. Thus, the corporate tax rate has not varied

Recent changes in tax rates havebeen more generous for high levels

of income than for low, withgreater effects on securities heldby upper-income groups, such as

common stocks and municipalbonds.

sufficiently to account for the significant variation inthe interest rate ratio.

Changes in the personal income tax rate sched-ule have been more dramatic. The 1981 ERTA re-duced the maximum personal income tax rate to 50percent, and reduced other personal income tax ratesby 25 percent over a three-year period. The TaxReform Act of 1986 dramatically reduced tax rates formost income levels. The maximum tax rate on per-sonal income was reduced from 50 percent to 38.5percent in 1987 and 33 percent for 1988 and lateryears. The 1986 Act also simplified the individualincome tax rate schedule by cutting the number ofbrackets from 14 to only four (15 percent, 28 percent,33 percent, and 28 percent), widening the incomelevels associated with each bracket and thus reducingthe problem of bracket creep.

The most recent change--the Revenue Reconcil-iation Act of 1990--further reduced the number of taxbrackets to three (15 percent, 28 percent, and 31percent) for 1991 and later years, thereby eliminatingthe "bubble" for upper-income levels. In addition,the amount of itemized deductions was reduced by 3percent of income over $100,000: if a taxpayer’s

18 These rates are for 1990 income. Connecticut has a 14percent maximum rate on interest and dividends; the Connecticuttax is levied only if 1990 federal adjusted gross income exceeds$54,000.


Major Tax Legislation in Recent Years

1976 Tax Reform Act ¯ Lengthened holding periods for long-term capital gains from sixmonths to one year.

¯ Created tax-sheltered Individual Retirement Accounts (IRAs).

1978 Revenue Act ¯ Widened personal income tax brackets.¯ Reduced maximum corporate income tax rate from 48% to 46%.

1981 Economic RecoveryTax Act (ERTA)

Cut maximum personal income tax rates from 70% to 50%.¯ Reduced personal income tax rates at all levels by 25% over a three-

year period.¯ Initiated indexation of tax brackets beginning in 1985.¯ Introduced super-accelerated depreciation of business assets (ACRS).¯ Expanded tax-sheltered investment opportunities

~troduced All Savers Certificates---expanded eligibility for IRAs--introduced net interest exclusion.

1982 Tax Equity and Fiscal ¯ Allowed deduction of only 80% of interest paid to carryResponsibility Act municipal bonds.(TEFRA) ¯ Established 10% withholding tax on interest and dividends.

1983 Social Security ActAmendment

¯ Subjected Social Security benefits to federal income tax.¯ Formula for Social Security benefit tax led to indirect taxation of

municipal interest.

1984 Deficit ReductionAct

Postponed ERTA’s interest exclusion.¯ Extended tax exemption for mortgage revenue bonds to 1987.¯ Reduced holding period for long-term capital gains to 6 months.. Limited depreciation on assets leased to tax-exempt entities.

1986 Tax Reform Act ¯ Reduced maximum personal income tax rate to 33% and maximumcorporate income tax rate to 34%.

. Replaced 14 personal income tax brackets with only four (15%, 28%,33%, 28%).

¯ Eliminated deduction of interest paid by commercial banks to carrymunicipal bonds.

. Placed limits on eligibility of "private-purpose" municipal bonds fortax exemption.

¯ Dramatically reduced the tax advantages of many "tax shelters,"such as real estate.

1990 RevenueReconciliation Act

¯ Established three tax brackets (15%, 28%, 31%).¯ Set maximum capital gains tax rate at 28%.. Reduced itemized deduction by 3% of income over $100,000,

resulting in an increase in the effective ordinary income tax rateby roughly 1%.


income is, say, $110,000, his itemized deductions arereduced by $300, effectively raising his marginal taxrate from 31 percent to about 32 percent.

Figure 5 reports the maximum statutory tax rateon personal income as well as the marginal taxbracket for four levels of real income between 1962and 1990. The maximum tax rate was 91 percent inthe early 1960s, then fell to 70 percent in 1965 whereit stayed through 1981 with a brief period of increase(1968-70) because of a surtax levied to finance theVietnam War. In 1982 the maximum rate was reducedto 50 percent, then to 33 percent in 1987.

The marginal tax rates for four levels of realincome, measured in 1980 dollars, are also shown inFigure 5. For all real income levels, the marginal taxbracket increased from 1962 through 1981; this in-crease was due to bracket creep. It is clear that thechanges in tax rates initiated by ERTA in 1981 andfollowed by the 1986 Tax Reform Act have been moregenerous for high levels of income than for low: themarginal tax rate at the $25,000 real income level wasactually higher in 1990 than in 1962. Thus, to theextent that tax rates affect investment decisions, smalleffects, if any, should be seen for securities heldby lower-income groups (bank deposits, corporatebonds) and the effects should be more dramatic forsecurities held by upper-income groups (commonstocks, municipal bonds).

Indexation of the Tax System

Prior to the 1980s, the tax system was basedsolely on nominal income levels. This meant that thedepreciation allowed businesses for plant and equip-ment was based on original cost, not replacementcost, and the tax rate schedule was based on themoney value of taxable income. This clearly pre-sented a problem in an inflationary environment suchas that prevailing in the 1970s and early 1980s.

One effect of inflation is that the cost of replacingphysical assets, such as vehicles, equipment, andstructures, exceeds the original cost of the assets. Asa result, historical cost depreciation--based on theoriginal cost of the depreciable asset--means thatbusinesses cannot fully recognize the cost of replac-ing equipment as a deductible cost of business. Theeffect is to overstate business income by including init the difference between "economic" depreciation ona replacement cost basis and historical cost deprecia-tion. This overstatement of business profits results inhigher tax payments and a deterioration of the firm’sfinancial position.

Figure 5

Marginal Personal Inco~ne Tax Ratesat Selected Income Levels (1980 Dollars)

lO0

90

80

70

60

50

40

30

20

lO

Percent

Maximum

$25,000

1 2 1966 1970 1974 1976 1980 1984 1988

Source: Internal Revenue Code and author’s calculations.

A second effect of an inflationary environment isthat it results in higher tax rates on real incomebecause individuals moved into higher tax bracketswhen their money income rose, even ,though theirreal incomes remained unchanged or even de-clined.19 This phenomenon, called "bracket creep,"raised the effective tax rate on real income over timeas inflation continued.

The 1981 Economic Recovery Tax Act introducedindexation of the personal income tax brackets, be-ginning in 1985. With indexation, the break pointsbetween tax brackets are increased according to ageneral price index. For example, in 1985 the tax ratewas 28 percent for taxable income in the range$31,120 to $36,630, but in 1986 the 28 percent rate wasapplied to taxable personal income between $32,270and $37,980, reflecting inflation of about 3.7 percentbetween 1985 and 1986. Indexation has eliminatedthe problem of bracket creep, at least for taxpayers

~9 This has been particularly important for personal incometaxes, which, for much of this century, had a complex set of taxbrackets. For example, in 1986, before the Tax Reform Act of 1986became effective, there were 14 tax brackets: for a married taxpayerfiling jointly, these ranged from 11 percent for income between$3,670 and $5,940 to 50 percent for taxable income over $175,250.


whose consumption bundle mimics the compositionof expenditures that the general price index repre-sents. For other taxpayers a weak connection be-tween real tax rates and inflation might remain, butthis has been reduced by the movement toward aflat-rate tax system in the 1986 and 1990 tax laws.

Other Aspects of the 1986 Act

As discussed in Section II, the 1986 Tax ReformAct imposed restrictions on private-acflvity use oftax-exempt bonds. This favored the tax-exempt bondmarket, forcing private borrowers back into the tax-able bond market and, to the extent that relativesupplies of taxable and tax-exempt debt affect relativeyields, raising the spread between taxable and tax-exempt yields.

A second piece of the 1986 legislation restrictedthe tax advantages of a wide range of "tax shelters."For example, depreciation of real estate was dramat-ically reduced, raising the effective tax rate on realestate investments. Tl~is elimination of tax shelterswas used as a selling point by the municipal bondindustry, the claim being that municipal bonds werethe sole remaining tax shelter. If this claim were true,the municipal bond rate should have fallen relative totaxable bond yields, reinforcing the effect of private-activity bond limitations.

A third non-tax-rate feature, also noted above,was the elimination of the commercial bank deduc-tion for costs of carrying municipal bonds. The effectof this change offset, to some extent, the effect of theother two legislative changes: municipal bond yieldswould rise relative to taxable bond yields as nonbanksectors were induced to increase their holdings ofmunicipal bonds.

Thus, the 1986 legislation not only dramaticallychanged the tax rate structure, it also introducedimportant non-tax-rate changes that could have, inprinciple, either favored or harmed the municipalbond market.

IV. The Determinants of Municipal BondYields

This section uses the insights developed in pre-vious sections as the basis for a statistical analysis ofmunicipal bond yields. The primary purpose is to testwhether the recent experience for the yields onmunicipal bonds is, in fact, consistent with the prop-ositions developed in the previous section. The anal-

ysis will focus on interest rate ratios, as defined in thefirst section and as shown in Figure 2.

The Traditional Explanation of the hlterest RateRatio

Why do interest rate ratios vary so much overtime? Why has the ratio been higher for long matu-rifles than for short maturities? Why has the differ-ence between the long and short maturities almostdisappeared in recent years? This section presents asimple model of the municipal bond market thataddresses these questions.

Assume that municipal bonds and taxable bondsare substitutes in investors’ portfolios. Each investorwill choose an amount of municipal bonds based onhis tax rate and on his assessments of the relativeliquidity of municipal bonds. For each investor, the.optimal holding of municipal bonds will be thatquantity for which (RM/RT) = A ÷ (1 - t), where t ishis tax rate and A is the liquidity premium required bythe investor. While the tax rate is exogenous to theinvestor’s decision, the liquidity premium is endoge-nous: as an investor contemplates increasing theamount he invests in municipal bonds, he will re-quire a higher interest rate ratio to compensate for theincreased risk and lower liquidity of municipalbonds.

The liquidity premium is due toseveral sources of risk inherent in

municipal bond ownership.

The liquidity premium is due to several sourcesof risk inherent in municipal bond ownership. Thefirst is lower liquidity and higher transactions costsdue to the serial form of municipal bonds: becauseeach bond is traded in separate strips, the averagetransaction is often small, which leads to increasedtransactions costs, and investors cannot sell theirmunicipal bonds as quickly as Treasury or corporatebonds can be sold. In addition, as shown in SectionII, the tax code penalizes municipal bonds relative totaxable bonds when bond prices fall. Finally, inves-tors are uncertain about future income tax rates andwill require some premium to compensate them forthis uncertainty. While the nature of each of these


risks is quite different, the term "liquidity premium"is used to describe the additional interest rate ratiorequired by investors to compensate for these extrarisks.

Assume that the liquidity premium is zero for thefirst dollar of municipal bonds held by an investor: ifan investor holds no municipals, he considers thefirst dollar of municipals to be equivalent to a dollar oftaxable bonds. This means that for intramarginalinvestors, the interest rate ratio will exceed the value(1 - t) by the liquidity premium required to inducethem to hold municipal bonds. But for the marginalinvestor, who holds a small amount of municipalbonds, the interest rate ratio is (1 - tin), where tm isthe marginal investor’s tax rate.

Figure 6 shows the demand functions for munic-ipal bonds of two investors, the "first investor,"whose tax rate, tmax, is the highest, and the "marginalinvestor," with tax rate tm. The quantity of municipalbonds acquired is measured along the horizontal axis.The vertical axis shows the interest rate ratio. Thehorizontal lines at interest rate ratios (1 - tmax) and(1 -- tm), respectively, show each investor’s demandfunction for municipal bonds if tax-exempt and tax-able bonds are perfect substitutes. In that case, inves-

tors do not require a liquidity premium and only taxrates matter in determining whether to buy a tax-exempt or a taxable bond. The upward-sloping solidlines labeled D1 and Dm are the actual demandfunctions, with the vertical distance to the broken linerepresenting the liquidity premium required to in-duce the investor to hold each quantity of municipalbonds.

Figure 6 assumes that the bond markets havesettled into an equilibrium in which the interest rateratio is just sufficient to induce a marginal investorwith tax rate tm to buy a small amount of tax-exemptbonds. The equilibrium interest rate ratio is (1 - tm),which is high enough to induce the first investor tohold QI* in tax-exempt bonds. For each investor, theinterest rate ratio is composed of two parts. The firstis the ratio required to give tax-exempts the sameafter-tax return as taxable bonds; for the first investorthis is (1 - tmax). The second part is the liquiditypremium required to compensate intramarginal in-vestors for the extra risks they attach, at the margin,to tax-exempt bonds. For the first investor the liquid-ity component is the value of ,~ at QI*, or I(Q~*) . Forthe marginal investor the liquidity component is (byassumption) zero.

RMRT

(1-tm)

(1-tmax)

Figure 6

h~dividual Investors h~ the Municipal Bond Market

First InvestorD1

(1 -tm)

Marginal Inv~....~ Dm

(Q~)Q1 Qm


Following an unfortunate convention, the term"windfall income" will be used to designate anyincome from tax-exempts in excess of the incomerequired to break even on an after-tax basis. For thefirst investor the dollar value of windfall income ismeasured by RT* (area A + area B).20 But RT* area Bis not really a windfall, for it is the amount of extraincome required to induce the investor to hold QI* ofmunicipal bonds. The only true excess income ismeasured by RT* area A. This is the "investor’ssurplus," which exists because the investor earnsinterest on his intramarginal investment in excess ofthe amount required. Note that in the case of a lineardemand function, the investor’s surplus will be 50percent of the investor’s windfall income.

Figure 7 shows the municipal bond market underthese conditions. The vertical axis represents theinterest rate ratio while the horizontal axis shows thequantity of municipal bonds outstanding. The up-ward-sloping schedule, marked DD, is the demandfunction for municipal bonds: as the interest rate ratiorises, more investors are induced by tax consider-ations to hold municipal bonds. As the market travelsup the demand schedule, the marginal tax rate ofinvestors is falling because new investors drawn intothe market have lower tax rates.

The bond supply schedule, SS, is assumed to bemoderately sensitive to the municipal bond rate,hence it is downward-sloping but with a steep slope.The bond supply function will also shift with the levelof the taxable bond rate. To understand this, notethat the supply of municipal bonds is affected by themunicipal bond rate, but the vertical axis in Figure 7 isthe interest rate ratio, which depends on both interestrates. As a result, the position of the supply schedulewill depend upon the level of the taxable bond rate: atany rate ratio, a higher RT implies a higher municipalbond rate and a lower quantity of tax-exempt bondsissued. Thus, SS will shift leftward (rightward) whenRT rises (falls). For expository purposes, we willassume that RT is at its equilibrium level and is notchanging.

This model suggests two basic demand-side de-terminants of the municipal bond yield. The first isthe maximum tax rate, tmax. A fall in tmax will makemunicipal bonds less attractive to the first investor,encouraging him to buy fewer municipals and shift-ing DD to the left.21 The second factor is the progres-sivity of the federal income tax schedule: a lessprogressive tax rate schedule will create a flatter slopeof the DD schedule so that the DD schedule rotatesclockwise at the rate ratio (1 - trnax). This creates a

Figure 7

The Market for Municipal Bonds

RMRT

1.0

(1 -tm)

(1-tmax)

S

’,S

D

QMOutstanding Stockof Municipal Bonds

lower rate ratio. Thus, the interest rate ratio shouldbe inversely related to the maximum tax rate anddirectly related to the degree of progressivity inincome tax rates. The econometric analysis in SectionV will support these hypotheses.

Before proceeding, it should be noted that inrecent years another school of thought on municipalbond yield determination has arisen. This school,associated with Fama (1977) and Miller (1977), arguesthat the personal income tax rate schedule--the focalpoint of the traditional explanation is irrelevant,and that the interest rate ratio is determined by thecorporate income tax rate (to). This "New View" isdescribed below in an appendix. Although it receivedsome empirical support in its early years, recentchanges in the structure of the market have weak-ened any validity of the New View.

20 Because the vertical axis is in units of taxable interest, anyarea in Figure 6 is measured in units of the taxable interest rate. Toconvert an area to a dollar value we must multiply it by RT. This iswhy the dollar value of windfall income is RT* (area A + area B),rather than simply area A plus area B.

21 Note that if taxables and tax-exempts were perfect substi-tutes, this would not be true. High-bracket investors would investall of their available funds in tax-exempts. A change in the tax ratethey face will alter the windfall income they receive, but not theamount invested.


The Ter~n Stn~cture of Municipal Bond Yields

The analysis of the previous sections assumes asingle type of municipal bond, and does not allowconsideration of such issues as the term structure ofmunicipal bond yields. The interest rate ratio rosesharply after 1986 for one-year and five-year terms,however, while it changed little for 20-year bonds.Figure 2 shows this clearly, and also shows that after1986 the one-year ratio was almost equal to thefive-year ratio.

The most widely held explanation of thesechanges in the term structure of municipal bondyields appeals to the notion of "market segmenta-tion." For much of this century commercial bankshave been important investors in the market fortax-exempt debt, and banks prefer (other thingsequal) to invest in securities with short to intermedi-ate maturities. Households, on the other hand, havetraditionally preferred longer-term bonds. This de-scription of the municipal bond market certainly fitthe data over the years when it was formed: the 1960sand 1970s.

However, as noted above, this picture changeddramatically in the mid 1980s when banks withdrewfrom investments in municipal securities. Becausebanks typically hold shorter maturities, the primaryimpact of this withdrawal was in those maturities.This meant that one or both of two things had tohappen. First, municipalities had to reduce theirissue of short- to intermediate-term debt; this couldbe accomplished by reducing capital outlays, by in-creasing use of tax revenues to finance capital out-lays, or by substituting longer-term debt for short- tointermediate-term bonds. Second, some nonbanksectors had to be induced to acquire short- to inter-mediate-term debt when they would not otherwisehave done so. Both of these adjustments require arise in yields on short- to intermediate-term bondsrelative to long-term yields.

The primary adjustment was of the second form:households responded by increasing their ownershipof municipal bonds, much of this in the short tointermediate maturities. The shortening of averagematurities held by the household sector was aided byseveral financial innovations that reduced the risks,and increased the liquidity, of municipal bonds.Chief among these was the formation of mutualfunds specializing in municipal debt of all maturities,but particularly of short to intermediate maturities.These allowed households with small portfolios todiversify their holdings as well as to gain liquidity by

check-writing privileges and by redemption of sharesat net asset value. A second innovation was thedevelopment of private firms providing municipalbond insurance. While bond insurance had been firstprovided in the early 1970s, the explosion of themarket for it in the 1980s was induced by the growingdominance of households in the municipal bondmarket.

V. An Econometric AnalysisIn this section a model of interest rate ratios will

be estimated. The variables to be explained are fourinterest rate ratios, the three shown in Figure 2 plusthe 10-year ratio, for the period June 1970 throughDecember 1989.

The model uses several tax rate variables tocapture the effect of changes in tax rate legislation. Torepresent the personal income tax rate schedule, themodel uses the ~naximum personal income tax rate(TMAX), and a measure of tax rate progressivity(PROGRSV), defined as the difference between themaximum personal tax rate and the tax rate for anindividual with $25,000 of real taxable income. Themaximum tax rate employed in defining these varia-bles is the rate paid by those in the highest incomebracket. Until 1987, this was also the highest ratelevied, but the 1986 Tax Reform Act introduced abubble in the tax rate schedule, so that the highest-income taxpayers paid less than the maximum taxrate. Thus, for 1988-90 the model uses 28 percent asthe maximum tax rate, not the 33 percent bubble rate.

The Traditional View of municipal bond yieldspredicts that the first tax variable (TMAX) shouldhave a negative coefficient, while the coefficient onthe second (PROGRSV) should be positive: a moreprogressive tax rate schedule, given the maximumrate, should increase municipal bond yields relativeto taxable bond yields. Note that because the tax ratedata are available annually, the same tax rate isassigned to each month in the year.

This econometric model does not incorporate theNew View, for two reasons. First, as noted above,evidence is abundant against the hypothesis thatcorporate income tax rates dominate the determina-tion of the interest rate ratios. Second, an experimentthat included the maximum corporate income tax ratedid not support the New View: the corporate incometax rate had the wrong sign and was not statisticallysignificant.

In order to capture the influence of anticipated

Septe~nber/October 1991 New England Econo~nic Review 31

future tax rates, several measures of the implicit futuretax rate were constructed: T2_5 is the implicit tax ratefor two to five years in the future, T5-10 is the implicittax rate five to ten years out, and T5_20 is the implicittax rate five to 20 years hence. Each is constructedfrom the yield curve data that were used to constructFigure 2, using the method described in Section Iabove. These constructed variables are available forevery month in the sample. Rather than include allthree in the regression for each rate ratio, only theones that seemed most relevant were included: t5_20in the 20-year regression, tsq0 in the 10-year regres-sion, and t2_s in the five-year regression. Anticipatedfuture tax rates were excluded from the one-yearregression because the relevant tax rate is known atthe time of purchase.

It was argued earlier that the anticipated variabilityof bond prices should affect the interest rate ratiobecause market discounts on municipal bonds aresubject to capital gains taxes, placing them at adisadvantage relative to taxable bonds if bond pricesfall, but giving them no advantage if bond prices rise.In order to reflect this, the Ibbotson Associates (1990)data on the monthly rate of change in prices oflong-term Treasury bonds were used to construct avolatility index for each month. The result is a vari-able, called VOLATILE, available for each month.22

Three dummy variables were used to capturefixed effects. One, labeled NY, is for the New YorkCity financial crisis, which is assumed to have oc-curred in the period June 1975 to December 1976.During this period, the yields on lower-rated munic-ipal bonds rose relative to yields on prime gradebonds, but the New York City crisis could haveaffected prime grade bonds as well, although thedirection of effect is not clear: NY could have apositive coefficient if the quality of high-grade bondswas called into question, or it could have a negativecoefficient if a flight to quality drove high-grade bondyields down.

A second dummy variable, named TRA86, ap-plies to the period January 1987 through the end ofthe sample. The 1986 Tax Reform Act had a variety ofeffects on the municipal bond market other that thosethat operate through tax rates (which are alreadycaptured in TMAX and PROGRSV). The limits onprivate-activity bonds and the severe limits on othertax shelters should induce a negative coefficient onTRA86, but the elimination of commercial bank de-ductibility of municipal carrying costs should have apositive effect. While the sign of the coefficient onTRA86 cannot be specified a priori, the common view

is that the elimination of carrying-cost deductibilitydominated the effect.

The third dummy variable, labeled Y86, refers tothe 12 months of 1986 and this is introduced tocapture any effects of the active and often-changingdebate about tax policy during that year.

Estimation of the model was done with Three-Stage Least Squares, using a correction for first-orderautocorrelation. Three-Stage Least Squares (3SLS) is amethod of joint estimation of a system of equationscombining Seemingly Unrelated Regression Equa-tions (SURE) and Two-Stage Least Squares (2SLS).The four interest rate ratios are viewed as a four-equation system because of the potential for omittedvariables common to interest rates at all four maturi-ties. The SURE method employs this information oncorrelations between residuals to derive efficient es-timates of the parameters.

The 2SLS aspect of 3SLS was necessary becausethe variables for anticipated future tax rates areendogenous: they are derived from the term structureof interest rates and, therefore, use the same interestrates that are used in defining the interest rate ratios.In order to eliminate this problem of feedback fromthe dependent variables to the variables T2_5, T5_10and T5_~0, a rather long list of instruments wasused.23

As noted above, the estimation was done with acorrection for first-order autocorrelation. The specificmethod of estimation involved two steps. First, eachof the four equations was estimated separately, using2SLS with an autocorrelation correction. This givesfour autocorrelation coefficients, one for each equa-tion; those autocorrelation coefficients are reported asthe variable RHO in Table 1. Second, the variables ineach equation were transformed to partial differencesusing the autocorrelation coefficient estimated forthat equation in the first stage, after which thetransformed variables were employed in a 3SLS esti-mation of the four-equation system.

22 The rate of change in bond prices is Ibbotson’s total returnon long-term government bonds in each month less the incomereturn due to coupons. The monthly volatility index is the squareof the deviation of the bond price change from its sample mean,divided by the sample variance. The variable VOLATILE is thesimple average of the volatility measure in the past three months.23 The instruments were all of the exogenous variables in thesystem, including a constant term, as well as the following addi-tional instruments, all available monthly: a time trend, the level ofthe CPI, the rate of inflation, the civilian unemployment rate, realpersonal income, the earnings-price ratio for the S&P500, and thethree-month Treasury bill rate.

32 Septetnber/October 1991 New England Economic Review

Table 1Three-Stage Least Squares Regression:Interest Rate Ratios, 1-5-10-20 YearMaturities, June 1970 to December 1989Dependent Variable (RM/RT)

Independent Maturity

Variable 20-Year 10-Year Five-Year One-Year

CONSTANT 0.3841 0.4636 0.3545 0.3275(42.53) (28.81) (44.35) (8.57)

T5_2o -.7004 n.a. n.a. n.a.(79.85)

T5_~o n.a. -.4832 n.a. n.a.(28.51)

T24 n.a. n.a. -.7705 n.a.(153.49)

VOLATILE .0011 .0019 .0014 .0070(1.46) (1.54) (1.96) (2.17)

TMAX -.0024 -.0043 -.0019 -.0083(3.27) (4.01) (2.69) (2.73)

PROGRSV .0017 .0031 .0015 .0071(2.42) (3.10) (2.38) (2.47)

NY -.0010 -.0031 -.0018 -.0089(.22) (.49) (.43) (.48)

Y86 .0258 .0475 .0183 .0780(4.60) (5.74) (3.44) (3.30)

TRA86 .0115 .0226 .0221 .1019(1.86) (2.50) (3.75) (3.92)

RHO .6088 .5170 .6310 .6023(2.83) (1.80) (5.47) (5.41)

.8818 .9526 .3071

1.9849 2.1597 2.1134

~2 .9514

DW 2.1845The method of estimation is 3SLS with correction for first-order auto-correlalion; RHO is the autocorrelation coefficient. The R2and DW arefor the transformed equations. The instruments are used for the fuluretax rate variables; see footnote 23 for instrument list. Numbers inparentheses are absolute value of t-statistics, corrected for instrumen-tal variables estimation.n.a. = not applicable.

The results, reported in Table 1, provide strongsupport for the insights gained from our discussion oftax legislation. With respect to the fixed effects, wefind that the New York City financial crisis played norole: the coefficient is negative in each equation,suggesting that this period was one of lower munic-ipal bond yields, but it is not statistically significant.

The dummy variable Y86 is positive and statisti-cally significant. The very active debate about taxpolicy in 1986 increased the interest rate ratios be-

cause of the uncertainty about future tax rates. Theimplied increase in the rate ratios in 1986 was about0.03 for 20-year bonds, 0.05 for 10-year bonds, 0.02for five-year bonds and 0.08 for one-year bonds. Atthe 1988-90 averages of Treasury bond yields, theserate ratio increases imply increases in municipal bondyields of 23 basis points (bp), 41 bp, 15 bp, and 62 bp,respectively.24

The variable TRA86 shows that the period ofeffectiveness of the 1986 Tax Reform Act (1987 andafter) was also a period of higher interest rate ratios.Like Y86, TRA86 appears to have the greatest impacton one-year bonds (which had an interest rate ratio0.10 higher after 1986), but it was both statisticallyand economically significant (roughly 0.01 to 0.02) forother maturities. This is predicted from the elimina-tion of the deduction for commercial bank carryingcosts. The municipal bond yield effects of TRA86,evaluated at 1988-90 Treasury bond yields, were 10bp for 20-year bonds, 20 bp for 10-year bonds, 19 bpfor five-year bonds, and 82 bp for one-year bonds.

The tax rate variables are all statistically signifi-cant with the signs predicted by theory. According tothe estimated coefficients, the reduction in the maxi-mum tax rate (TMAX) from 50 percent to 28 percentafter the 1986 Tax Reform Act reduced the interestrate ratios by about 0.05 for 20-year bonds, 0.10 for10-year bonds, 0.04 for five-year bonds and 0.18 forone-year bonds. The corresponding declines in mu-nicipal bond yields, evaluated at 1988-90 Treasurybond yield levels, were 46 bp, 82 bp, 35 bp, and 146bp, respectively.

The tax progressivity variable (PROGRSV) also isstatistically and economically significant in eachequation. The 1986 Tax Reform Act reduced theprogressivity variable by 22 percentage points, from22 percent in 1985 to zero in 1988. This reduced therate ratios by 0.04 for 20-year bonds, 0.07 for 10-yearbonds, 0.03 for five-year bonds and 0.15 for one-yearbonds. The corresponding decreases in municipalbond yields are 32 bp, 57 bp, 27 bp, and 120 bp.

In each equation the coefficient on anticipatedfuture tax rates is negative and statistically signifi-cant: when the market expects tax rates to be high,the interest rate ratio is reduced as tax-exempt bondsare substituted for taxable bonds. The very hight-statistics attest to the statistical importance of tax

24 The 1988-90 averages for Treasury bond yields were 8.7percent for 20 years, 8.6 percent for 10 years, 8.5 percent for fiveyears, and 8.0 percent for one year.

September/October 1991 New England Economic Reviezo 33

rate anticipations, and the size of the coefficientsattests to their economic significance.

In order to assess the economic significance ofthe 1986 Tax Reform Act, the change in the interestrate ratio between 1982-85 and 1988-90 attributable tochanges in statutory tax rates (TMAX and PROGRSV)was calculated, as well as the fixed effect captured inthe dummy variable TRA86. The changes in rateratios were converted to changes in municipal bondyields, using the 1988-90 average values of the fourTreasury bond yields.

The results are reported in the top portion ofTable 2. These three tax policy variables account for arise in the municipal bond yields by 25 to 44 bp for thelonger maturities, and by a large 107 basis points forthe one-year maturity. The primary source of thelarge increase at the one-year maturity is the TRA86dummy variable. That this is most important at the

Table 2Effects of Tax Policy on Interest RateRatios and Municipal Bond Yields,1982-85/1988-90Independent Maturity

Variable 20-Year 10-Year Five-Year One-Year

Tax Policy Variables

TMAX +.0528 +.0946 +.0418 +.1826PROGRSV -.0361 -.0659 -.0319 -.1509TRA86 +.0115 +.0226 +.0221 +.1019

Subtotal ratio +.0282 +.0513 +.0320 +.1336RM +25 bp +44 bp +27 bp +107 bp

short maturity is consistent with the withdrawal of aprimary short-term lender--commercial banks--fromthe tax-exempt debt market.

Another avenue for tax policy is anticipated taxrates. The bottom portion of Table 2 assesses theeffects of the anticipations variables after the TaxReform Act of 1986. The results suggest only a minoreffect of tax rate anticipations for 20-year and 10-yearbonds: the implied yield changes are -11 bp and -7bp, respectively. However, a 52 bp increase occurredin five-year yields in the 1987-89 period. This sug-gests that in this period tax rates were expected todecline further.

The combined effect of all four tax policy varia-bles (Table 2) is to increase municipal bond yieldsafter the 1986 Tax Reform Act, with the magnitudebeing greater for shorter maturities. The impacts aregreater for shorter-term bonds, ranging from a mild14-basis-point increase for 20-year bonds to a 107-basis-point increase for one-year bonds.

Bond price volatility (VOLATILE) increases therate ratios, as the earlier discussion of capital gainstaxation suggests. While the t-statistics suggest thatthis effect is not as reliable as the effect of taxvariables, the evidence does support the hypothesisthat the more volatile are bond prices (hence thehigher the probability of a price decrease), the lessdesirable are municipal bonds. This is the effect ofcapital gains taxes on the relative desirability oftax-exempt and taxable bonds when market dis-counts exist.

Thus, the data provide very strong support forthe role of personal income tax legislation, and theelimination of bank deductibility of carrying costs, inaffecting interest rate ratios.

Anticipated Tax Rate Variables

T5_2o -.0123 n.a. n.a. n.a.T,~.I o n.a. -.0076 n.a. n.a.T2_5 n.a. n.a. +.0616 n.a.

Subtotal ratio -.0123 -.0076 +.0616 n.a.RM -11 bp -7bp +52bp n.a.

Total ratio +.0159 +.0437 +.0936 +.1336RM +14 bp +37 bp +79 bp +107 bp

Source: Coefficients in Table 1 times average values of variables in1982-85 and 1988-90. The conversion from interest rate ratios tomunicipal bond yields is done at the 1988-90 average for Treasurybond yields: 8.75 percent, 8.63 percent, 8.45 percent, and 8.01percent, respectively.n.a. = not applicable.bp = basis points.

VI. SummaryThis article reviews the performance and chang-

ing structure of the municipal bond market in recentyears. It shows that the ratio of yield to maturity onmunicipal bonds to yields on U.S. Treasury bonds(the interest rate ratio) has varied greatly in the lasttwo decades, that the interest rate ratio is greater forlonger maturities, that the yield curve for municipalbonds has sloped upward more rapidly than the yieldcurve for Treasury bonds, and that the yield curveshapes for municipals and Treasuries became muchmore similar after the mid 1980s.

Dramatic changes have occurred in the characterof municipal bonds outstanding, with private-activity

34 September/October 1991 New England Economic Reviezo

bonds rapidly gaining market share in the 1970s andearly 1980s, and losing market share just as rapidlyafter tlqe mid 1980s. In addition, major changes inmunicipal bond ownership occurred, primarily thedramatic withdrawal of commercial banks from themarket after the mid 1980s, with households takingthe place of banks as the dominant investors.

The paper reviews the main aspects of the Inter-nal Revenue Code that affect the demand for munic-ipal bonds. It argues that municipal bond yields willrise relative to Treasury bond yields when personalincome tax rates fall or become more progressive, andthat the exposure of municipal bonds to capital gainstaxes when market discounts emerge will place mu-nicipal bonds at a disadvantage in periods whenbond prices are more volatile. Furthermore, it isargued that anticipated future income tax rates willhave a strong impact on municipal bond yields.

Recent tax legislation was examined in somedetail, particularly the 1986 Tax Reform Act. That Actnot only radically revised the income tax rate sched-ule-moving toward a flat-rate system with fewbrackets~but it also contained non-tax-rate legisla-tion that dramatically affected bond yields. Of partic-ular importance is the elimination of commercial bankdeductions for municipal bond carrying costs, whichshould raise the interest rate ratio. Offsetting this,perhaps only partially, were the Act’s severe restric-

tions on tax-sheltered investments, which made mu-nicipal bonds more attractive.

The paper concludes with an econometric analysisof the interest rate ratio for four maturities: one-year,five-year, 10-year and 20-year bonds. The results of thisstatistical analysis provide strong support for the im-portance of tax legislation in general, and for the prop-ositions outlined above about the role of personalincome taxes and the 1986 Tax Reform Act. In particu-lar, the analysis finds that both the decrease in themaximum personal income tax rate, and the decline inthe progressivity of the tax rate schedule, had effectsthat were statistically and economically significant, thatthe 1986 Tax Reform Act had a positive fixed effect onmunicipal bond yields, and that tax rate anticipationswere very important.

While each of these factors was statistically andeconomically significant, they tended to have offset-ting effects. The combined effect of these tax policyvariables after 1986 ranged from a small 14-basis-point increase for 20-year bonds to a large 107 bpincrease for one-year bonds.

The results also show that the debate during 1986about the 1986 Tax Reform Act--as opposed to itsactual implementation--increased interest rate ratios.This suggests independent evidence for the impor-tance of anticipated future tax rates.

Appendix: The "’New View" of Municipal Bond Yields

The "New View" of municipal bond yield determina-tion can be understood in two ways. Fama (1977) arguedthat corporations, primarily commercial banks, are themarginal investors in municipal bonds. Thus, the DDschedule follows the personal tax rate schedule for lowquantities of municipal bonds outstanding; for these quan-tities, individuals with tax rates between tma× and tc willprovide the funds. However, the DD schedule becomeshorizontal at a ratio of (1 - to): any tendency for the ratio togo above that level will induce banks to enter the market insufficient volume to restore the ratio to (1 - to). Because thevolume of municipal bonds outstanding is great enough tofully absorb the funds of high-bracket investors, commer-cial banks become the marginal investors and the equilib-rium rate ratio will be RM/RT = (1 - to).

Miller’s (1977) explanation of the New View is slightlymore exotic. According to the stripped-down Miller ver-sion, individual investors~and the personal tax rate sched-ule~determine the demand function, so DD remains theeffective demand schedule. But Miller argues that theeffective supply schedule for municipal bonds is horizontalat the interest rate ratio (1 - to).

In Miller’s view, common stocks and municipal bondsare perfect substitutes in investors’ portfolios, and commonstocks are virtually tax-exempt.2~ As a result, in equilibriumthe equality RM = Rs will hold, where Rs is the yield oncommon stocks. Because a corporation will choose itsdebt-equity ratio so as to minimize its cost of capital, if therequired return on stocks (Rs) exceeds the after-tax cost ofdebt (RT), the corporation will sell bonds; this occurs whenRs/RT > (1 - to). On the other hand, if the cost of equity isless than the after-tax cost of debt, corporations will sellequity; this occurs when Rs/RT < (1 - to). Since all corpo-rations face the same tax rate, all will finance themselveseither with debt or with equity unless Rs/RT = (1 - t~), inwhich case each firm will be indifferent; some firms willchoose debt, others will choose equity, and still others willfinance themselves with both debt and equity. This de-

2s Because dividend-paying stocks can be held by low-taxsectors (such as pension funds) and capital gains taxes can bedeferred, Miller’s assumption that common stocks are tax-exempthas some merit.


scribes the equilibi:ium. Because equity and municipalbonds are (according to Miller) perfect substitutes in inves-tors’ portfolios, so Rs = RM, the debt-equity decisions ofcorporations will ensure that RM/RT = (1 - to): the interestrate ratio is determined by the corporate income tax rate. Ineffect, the supply schedule SS is horizontal at the interestrate ratio (1 - to). At any higher ratio municipalities will notalter their debt decision, but "corporations will sell lessequity that, under the Miller assumptions, is equivalent toless tax-exempt debt.

The New View received some empirical support in itsearly years, a prominent example being the work of Trc-zinka (1982). However, it fails to fit the 1980s data, asshown by Fortune (1988) who used Trczinka’s method, andby Peek and Wilcox (1986), who used a different method.

Furthermore, the event analyses by Poterba (1986, 1989)have demonstrated the importance of the personal incometax. A more recent study of the importance of personalincome tax rates is Fortune (1991b).

In addition, recent changes in the structure of themarket clearly have weakened any validity of the NewView. For example, Fama’s explanation is based on highcommercial bank participation in the municipal bond mar-ket, but banks have been notoriously absent from themarket for municipal bonds in the 1980s. Also, Miller’sexplanation is less convincing in the 1980s, when corpora-tions clearly were not balancing debt and equity at themargin, but were apparently at a corner solution, issuingdebt in large quantities and retiring equity.

Refere~lces

Bushman, Charles and Philip S. Winterer. 1983. "Legal Arbitrage."In Fabozzi, Frank J. and others, The Municipal Bond Handbook,Volume I, Chapter 15, pp. 204-27. New York: Dow Jones-Irwin.

Fama, Eugene. "A Pricing Model for the Municipal Bond Market."Unpublished working paper, University of Chicago (May).

Fortune, Peter. 1973. "Tax Exemption of State and Local InterestPayments: an Economic Analysis of the Issues and an Alterna-five." Federal Reserve Bank of Boston, New England EconomicReview, May/June, pp. 3-31.

¯ 1988. "Municipal Bond Yields: Whose Tax Rates Matter?"National Tax Journal, vol. 41, no. 2, (June), pp. 219-33.

¯ 1991a. "Stock Market Efficiency: An Autopsy." FederalReserve Bank of Boston, New England Economic Review, March!April, pp. 17-39.

¯1991b. "Do Bond Yields Forecast Tax Policy?" Tufts Uni-versity Working Paper No. 9. Medford, MA: Tufts University.

Huefner, Robert. 1971. Taxable Alternatives to Municipal Bonds: AnAnalysis of the Issues. Research Report No. 53, Federal ReserveBank of Boston, Boston, MA.

Ibbotson Associates. 1990. Stocks, Bonds, Bills and Inflation Yearbook.Chicago, IL: Ibbotson Associates.

Kenyon, Daphne. 1991. "Effects of Federal Volume Caps on Stateand Local Borrowing." National Tax Journal (forthcoming).

Kidwell, David S., Timothy W. Koch and Duane R. Stock. 1984. "The

Impact of State Income Taxes on Municipal Borrowing Costs."National Tax Journal, vol. 37, no. 4, (December), pp. 551-61.

Miller, Merton. 1977. "Debt and Taxes." Journal of Finance, vol. 32,no. 2, (May), pp. 261-75.

Pechman, Joseph A. 1987. Federal Tax Policy. Washington, D. C. :The Brookings Institution.

Peek, Joe and James Wilcox. 1986. "Tax Rates and Interest Rates onTax-Exempt Securities." Federal Reserve Bank of Boston, NewEngland Economic Review, January/February, pp. 29-41.

Petersen, John E. 1991. "Innovations in the Tax Exempt Market."National Tax Journal (forthcoming).

Poterba, James M. 1986. "Explaining the Yield Spread betweenTaxable and Tax-Exempt Bonds: The Role of Expected TaxPolicy." In Rosen, Harvey S., ed., Studies in State and Local PublicFinance, pp. 5-49. Chicago, IL: University of Chicago Press.

¯1989. "Tax Reform and the Market for Tax-Exempt Debt."Regional Science and Urban Economics, August, pp. 537-62.

Trczinka, Charles. 1982. "The Pricing of Tax-Exempt Bonds andthe Miller Hypothesis." Journal of Finance, vol. 37, no. 4, (Sep-tember), pp. 907-23.

Zimmerman, Dennis. 1991. The Private Use of Tax-Exempt Bonds:Controlling Public Subsidy of Private Activity. WashingtonD. C.: The Urban Institute.

36 SeptemberlOctober 1991 New England Economic Review

long-term bonds address the

Documents