Financial innovations in municipal securities markets

Stephan David Whitaker (Research Department, Federal Reserve Bank of Cleveland, Cleveland, Ohio, USA)

Journal of Public Budgeting, Accounting & Financial Management

ISSN: 1096-3367

Publication date: 3 September 2018

Abstract

Purpose

The purpose of this paper is to measure how frequently innovative financial products appeared or became widely adopted in the municipal securities markets over the last two decades; and also investigate what types of issuers adopted the innovations, the relationship between yields and innovation and the patterns of diffusion within states.

Design/methodology/approach

Using comprehensive data on municipal securities issued from 1992 to 2015, the author searches for financial innovations as defined in the literature. The author uses issuer fixed effects models to characterize the relationship between yields and use of innovative products. Other models provide estimates of the conditional correlations between issuer characteristics and innovation usage. Finally, the author fits trend models to identify significant differences in the pace of adoption between different types of issuers.

Findings

In total, 35 security features fit one or more definitions of innovation. Extensive analysis is presented for four innovations that represent significant transfers of risk: variable rates, put options, corporate backers and derivatives. Small issuers used these innovative products, but the largest issuers adopted them to a greater extent. Usage appears to diffuse within states. Issuance of innovative securities fell during the financial crisis and has not recovered. Novel securities since the financial crisis have been created by legislation rather than by market participants.

Research limitations/implications

The data appear to cover all or nearly all municipal securities, but they do not cover loans or other types of municipal borrowing.

Practical implications

This analysis reveals that financial innovations in municipal securities markets usually take the form of a rare practice becoming widespread rather than a never-before-seen feature appearing in the market. Changes in response to legislation are an exception.

Social implications

Regulators concerned about financial stability can monitor the expansion of formerly rare securities features. This will be informative about new risks or transfers of risk in the market. They can also anticipate that expanded use of an innovation by states and high-volume issuers will be followed by adoption of the innovations by smaller, less sophisticated issuers in subsequent years.

Originality/value

This paper is the first attempt to empirically analyze the extent of financial innovation in municipal securities. Existing public finance literature has proposed definitions of financial innovation, qualitatively documented some specific innovations and empirically analyzed others. However, no previous study has empirically analyzed the entire municipal securities market for all possible innovations.

Keywords

Citation

Whitaker, S. (2018), "Financial innovations in municipal securities markets", Journal of Public Budgeting, Accounting & Financial Management, Vol. 30 No. 3, pp. 286-314. https://doi.org/10.1108/JPBAFM-02-2018-0006

Download as .RIS

Publisher

:

Emerald Publishing Limited

Copyright © 2018, Emerald Publishing Limited


1. Introduction

In a variety of articles published over the last several decades, researchers have attempted to define financial innovations in municipal securities markets. The literature has asserted that there has been considerable innovation in this area. However, the evidence presented has taken the form of identifying the many different types of securities that have appeared. Other papers have explored individual innovations in isolation. This study represents one of the first attempts to systematically characterize the extent and intensity of innovations in municipal securities. I will attempt to answer questions such as: how many types of securities can be considered innovative according to the definitions in the literature? How much market share did these innovative products capture? Additionally, which types of municipal issuers led the adoption of the innovations?

Using the Mergent Municipal Bond Securities Database, this study identifies financial innovations that were rarely used in the 1990s but their market share expanded immensely through 2008. After 2008, their use has plummeted back to previous levels. At the peak of the last business cycle, over 39 percent of municipal debt was issued using securities with features that were uncommon during the previous business cycle.

Using a variety of definitions from the literature, it is possible to identify 35 innovations among the municipal securities tracked in the Mergent database. Of these, four are particularly interesting because they entail substantial transfers of risk among market participants. These are municipal securities with variable rates, put options, corporate backers (in addition to the municipal issuers and financial intermediaries) and derivative structures. From this exercise, we learn that it is unusual for entirely new municipal finance innovations to appear unless they are created by federal legislation. However, popular innovations arise when security features that have been rare for many years begin to rapidly expand their market share and become widely used.

After linking the municipal securities data to the Census of Governments (COG) data, we can look at the characteristics and scale of the issuers that adopted the innovations. We can see that state governments and very large cities, counties and special districts were the most rapid adopters of the innovations in terms of market share. The thousands of small cities, counties, and districts also experimented with the innovative products, but to a lesser extent. There is evidence that use of innovations is positively related to use by issuers in the same state in recent years.

Understanding the frequency and nature of innovations’ appearance is valuable because innovations can create both new opportunities and new risks. New products can enable borrowers and investors to do something more efficiently or to do something they previously could not do (Finnerty and Emery, 2002). Innovations, by definition, do not have a historical record that can be used to price their risk, so there is a tendency for market participants to overuse them before they are fully understood (Gennaioli et al., 2012; Beck et al., 2016). In instances in which an innovation’s benefits outweigh its costs, we might seek policy interventions that help borrowers to access the innovation. If risks are not fully understood or reflected in prices, the proper policy might be oversight or limitation of the innovative product. Both policies rely on an understanding of when and how innovations emerge. Studying the patterns of previous municipal innovations, as in this study, provides the best guidance about how future products are likely to be adopted.

The recession that began 2008 is distinguished from the preceding 12 recessions by the central role played by the finance industry and new financial instruments (Financial Crisis Inquiry Commission, 2011). The crisis followed a 25-year period which had witnessed a proliferation of financial innovations. It is reasonable to expect that innovations were also proliferating in the realm of municipal securities or transferred in from corporate debt practices. However, the opposite prior expectation, that innovations were not being created and popularized in municipal markets, is also plausible. Municipal debt is regarded by many as a large but routinized market where innovation is not valued or allowed (Lebenthal et al., 2006)[1].

The literature has not been in a position to distinguish between these two possibilities because past studies have either listed securities or features without quantifying their use (Hildreth and Zorn, 2005) or empirically investigated a single innovation (Stewart and Cox, 2008). The effort required to collect data from surveys or financial disclosure documents has forced researchers to focus on only the largest issuers and on single innovations at a time. Since innovations that no one has investigated could exist, these studies could not speak to how many innovations are in the market or how widespread such innovations have become. In this study, a national and comprehensive data set is allowed to identify any innovation recorded, rather than selecting them ex ante. The data also make it possible to observe the issuance of infrequent issuers and compare it to that of frequent, high-volume issuers.

This paper proceeds as follows. Section 2 reviews the relevant literature on financial innovation in the corporate and municipal sectors. Section 3 describes the data sets, their coverage and their ability to measure innovations. Section 4 discusses the innovations that can be found in the data and gives a brief background on the most interesting ones. Section 5 investigates the patterns of the adoption of the innovations. Section 6 concludes with a brief discussion of the findings and implications.

2. Literature

The literature on financial innovation in corporate securities has both a long history and considerable breadth, while work on financial innovation in municipal securities is much more limited. In this section, I will briefly review articles that define the term “innovations” and seek to identify financial instruments that meet their criteria.

Although the literature on financial innovation dates back further, a good starting point for this analysis is a 1986 article by Merton Miller. The article looks back over innovations in corporate finance that had occurred since 1966. To define innovation, Miller (1986) invokes the concept of the statistical test for time series breaks, in which a trend diverges from a path predicted by the history before the break. A financial product is an innovation when it is significantly different from what preceded it. The analogy remains conceptual because Miller does not propose a measure of how different a new product is from the most closely related precedent. He does provide many examples of innovations, but they are selected subjectively.

Miller also discusses what changed in the environment that brought existent but obscure products into a new, prominent position in the market. He argues regulations and taxes are the primary motivations for financial innovations. If different rates of taxation are applied to different types of income, securities will be devised to transmute the higher-taxed income into the lower-taxed type. Miller defines “successful” innovations as those that save their users tax payments at least until the tax code is changed. “Significant” innovations are those that are not transitory, but continue to grow after the rule that initially motivated them has gone away. Eurodollar, Eurobond, zero-coupon bonds and swaps are given as examples of significant innovations. In this analysis, I will be able to identify four innovations in the municipal debt markets that fit Miller’s definition of “successful.” For several years at least, these products offered issuers and investors desirable terms. Whether the innovations survived the changing of the tax or regulation which prompted them is less clear. After the financial crisis, they continue to be used at diminished levels. One of the innovations was being represented by sellers as a short-term security and was treated as such on the buyers’ balance sheets. Regulators had to force the market to recognize the securities as long-term investments (Stewart and Smith, 2012). That change, along with other regulatory changes following the financial crisis, may be discouraging use of some of the innovations, but the low interest rate environment is probably the larger deterrent to a robust reemergence in the market.

Going beyond Miller’s examples, Finnerty (1988) compiled a list of corporate securities innovations. The list was updated in Finnerty (1992) and Finnerty and Emery (2002). Finnerty and Emery employ a slightly different definition of innovation than Miller, stating that an innovative product is one that “enable(s) market participants to either accomplish something more efficiently, or accomplish something they could not accomplish previously.” Their indicator of something being “truly innovative” is persistence. Their assumption is that if an apparent innovation is created only to enrich the investment bank or financial service provider, clients will eventually discover the lack of value and stop demanding the product. Their definition is comparable to Miller’s “significant innovation” definition in that the innovation has to continue after changes are enacted to the tax or regulation that initially sparked the expansion of the innovation. Finnerty and Emery would say that products that disappear after a tax change are not innovations at all, while Miller would call them successful innovations but not significant innovations. Finnerty and Emery categorize the corporate finance innovations they identify into those related to debt, preferred stock, convertible securities and common equity. Among the 48 debt innovations they list, 12 feature some type of variable interest rate. Variable rates are found to be central to the innovations in municipal markets examined in this study.

In an empirical study that followed the qualitative discussions, Tufano (1989) used data from investment banks to demonstrate that the originator of a financial innovation could profit from the innovation, even after competitors imitated the product. This first-mover advantage could serve as a motivation for developing innovations. For the purposes of Tufano’s paper, the definition of an innovation was a corporate or mortgage-backed security that was first traded between 1974 and 1986 and could be observed in the data after its introduction. The analysis presented here is similar to Tufano’s in that it depends on a commercial data set to measure the existence of the product. It is possible that the innovations existed earlier but were not reflected in the data sets he uses. If a product exists that has never been recorded in the data, it could not be identified as an innovation.

Fourteen years later, Tufano (2003) provided a broad overview of the financial innovation literature in a handbook chapter. He included a slightly broader definition of innovation, characterizing it as “the act of creating and then popularizing new financial instruments” (p. 310). Notably, this definition requires only popularization, not persistence. As in the 1989 article, the act of creating could be identified in the data if one observes an extended period in which a type of security is never recorded, followed by a unique first appearance. In this analysis, I will identify both first appearances of innovations and several popularizations.

The works highlighted above are some that provided definitions of “innovation.” A sizable literature that researched aspects of specific innovations was developed over the intervening years. In 2004, Frame and White published a review of this literature. They classify the topics of empirical studies into four types: conditions that encourage innovation, users of innovative products, diffusion and profitability and social welfare consequences. Within the category of papers Frame and White label “users of financial innovations,” they include two papers regarding the issuers of innovative securities (Alderson and Fraser, 1993; Goyal et al., 1998). The “users” category would be the most appropriate of Frame and White’s classifications for the analysis presented here. I make use of information on the securities’ issuers rather than on the financial intermediaries or investors.

Frame and White argued that as of 2004, few papers had been published that had empirically tested hypotheses related to financial innovations. The analysis below does include some regressions that can speak to hypotheses one may have regarding what type of issuer is likely to adopt financial innovations more rapidly or with a greater share of their issuance volume.

While the papers discussed above all focus on innovations in corporate or consumer finance, there are at least some publications that address innovation in the municipal finance field[2]. Kidwell and Rogwski (1983) studied the interest cost savings realized through bond banks and discussed the adoption of bond banks by smaller issuers. In 2005, Hildreth and Zorn published an overview of what they referred to as the “evolution” of municipal debt markets. The time period they considered was 1980 through 2005, and they highlighted the Tax Reform Act of 1986 as the cause of many of the changes that occurred. They argue that volume limitations and other restrictions placed on private activity bonds concentrated more of the market’s changes in that area. Innovations have also been driven by an issuers’ desire to work around the federal limitations on refinancing, to circumvent state limitation on indebtedness and to market terms attractive to investors.

Hildreth and Zorn’s paper features a section on innovations in debt instruments and provides a lengthy list of examples, including, “variable-rate obligations, derivative securities, interest rate swaps, tax-exempt inverse floaters, puts, warrants, residual interest bonds, and zero coupon bonds.” The authors describe variable rate demand bonds and auction rate securities (ARS) in detail as examples. The paper, which is primarily narrative, does not quantify the growth of the innovative debt instruments they describe. In hindsight, the article was written in the middle of the largest expansion of the products the authors mention. Hildreth and Zorn do not provide a formal definition of innovation so it is possible they consider some instruments to be innovations even if the instrument remains rare. That would be a different definition than those discussed above, all of which required that the innovation become popular at some point. However, the authors cite papers from the early 1980s that discuss the innovations, suggesting many of the innovations were in existence before 1981. In this analysis, we can measure the frequency and par values associated with many of the innovations Hildreth and Zorn mention. We can observe the innovations’ growth leading up to when their paper was written along with the products’ subsequent declines.

3. Data and variable definitions

The data used in the analysis originate in the Mergent Municipal Bond Securities Database, and they are supplemented with the COG. The Mergent database contains the characteristics of over 3m securities issued from 1992 to 2015[3]. An accompanying file groups these bonds into approximately 340,000 issue series and provides the name of the issuer. The Mergent data are merged with the COG data using the issuer name. The Mergent data include 65 variables that can be mapped into 177 potential innovations for municipal securities. Each category of the categorical fields was considered as a potential innovation, as were extreme values of the continuous variables.

In conducting a study of innovation using a commercial data set, there is the central concern of whether the data set is accurately reflecting new types of securities and new features. The data are used by investors, financial advisors, underwriters and other industry professionals for statistical analysis and to find the Committee on Uniform Security Identification Procedures numbers (CUSIPs) of securities with specific characteristics. From the 3m CUSIPs, a data user may wish to search for those issued in a specific state and year, and with a specific feature. When a new feature appears on a security, Mergent may anticipate that users will want to be able to search for securities that use the new feature. It is also possible that the feature is added after customers request the ability to track it, which could delay the tracking until market participants recognize the importance of the feature. If needed, Mergent creates a new category in an existing categorical field. We can observe this in the data in the addition of certificates of participation, tobacco bonds, and several other debt types. In some cases, such as the addition of a new call option, Mergent creates a new flag variable. The make-whole call flag first appears in 1997, for example.

In aggregate, the Mergent data are comparable to other published sources[4]. The most commonly cited source of figures for total annual issuance is Thomson Reuters SDC[5]. From 1996 to 2012, the Mergent totals are always within 9 percent of the issuance totals reported by Thomson Reuters SDC. The Mergent total is higher in most years but drops below the Thomson Reuters SDC totals in 2013 and 2014. The correlation between the Mergent and Thomson Reuters series is 0.98. Another data source is the Census Bureau’s COG Finances and Annual Survey of Local Government Finances. The Bureau reports an estimate of “long-term debt issued.” Comparison of this data and the Mergent data is more challenging because the Census reports totals for the jurisdictions’ most recent fiscal year rather than calendar year[6]. Fiscal years vary, but we can aggregate issuance in the Mergent data from July through June as a closer approximation to fiscal years. Calculated this way, the Mergent and Census series have a correlation of 0.93 between 1993 and 2014. The grand total of par value issued over the study period is 10.8 percent higher in the Mergent data than the grand total reported in the Census data. It is reassuring that the Mergent totals are usually above the totals in other sources. That suggests Mergent’s criteria for inclusion are the broadest, so it is the most likely to include innovative securities that the other collections are not yet recognizing.

To increase our confidence in the Mergent data, I created a random sample of 120 CUSIPs and searched for their official statements. In total, 104 of the official statements were available online, and I reviewed them to confirm consistency with data items in the Mergent database. Out of 7,253 features I could find reference to in the official statements, I identified only 61 instances of the official statement’s either contradicting the Mergent data or providing information that was missing from the Mergent data. The fields with the most inaccuracies were the offering type (negotiated or competitive, 16 inaccuracies), the count of financial advisors involved in the transaction (11 inaccuracies), and the designation of the bond insurance provider (8 inaccuracies). Measures based on these three variables are excluded from the main analysis because of the inaccuracies discovered. Other than the offering type (negotiated or competitive), I did not find any instances of a feature that would appear to be an innovation because it is recorded with increasing accuracy in later years.

The supplemental microdata from the COG are collected every five years, and this analysis uses the 1992, 1997, 2002, 2007 and 2012 censuses. The COG enumerates every subnational government in the USA including states, counties, cities (or towns), special districts and school districts. The COG data are merged with the Mergent data by extracting the government type and name from the “issuer long name” variable and matching them to the government type and name variables in the COG[7].

4. Identifying innovations

In the literature reviewed in Section 2, there are articles in which authors provide examples of innovations (Miller, 1986; Hildreth and Zorn, 2005). While their process of selecting examples is difficult to reproduce, other articles applied criteria that can be defined for this analysis. Finnerty and Tufano suggest innovations may be identified if they make an initial appearance in the data, are popularized and persist. To apply these criteria, specific thresholds must be chosen. One must decide how many time periods (month, quarters or years) must pass from the beginning of the data coverage before the first appearance of a feature can be treated as its creation[8]. To declare an innovation has been popularized, one can chose a market share threshold. Persistence could be measured by observing that the innovation remains above the popularization threshold, or that the innovation continues to be observed in the later years of the data (a threshold of zero). In the remainder of this section, I will place each of the 177 features measured in the Mergent data into one of seven categories based on the pattern of its market share. The criteria and designations are summarized in Table I. If innovations are simply features that did not exist and then started to be used, then 18 features could be call innovations. Using a popularization criteria, the 17 features that have minor usage in the first years of the data and then grow to be widely popular have a strong case for being considered innovations. Of these 35 features, 22 persist with at least modest market shares. For further analysis in Section 5, I select four of these innovations that appear to be consequential for risk and pricing. The four were also among those with the largest increases in their market share, which enables them to be informative about the adoption of innovations in general.

4.1 First-appearance innovations

To identify a first appearance in the data, I selected features that where not used at all in 1992 or 1993. Requiring multiple years of absence from the data, as opposed to multiple months or quarters, avoids misidentifying a feature that is more likely to be used in certain parts of the fiscal year. Using a first-appearance criteria, 18 of the 177 features could potentially be innovations because they do not appear at all in the first two years of the data, and then they start to appear in later years. The first-appearance items include several that were created by the American Recovery and Reinvestment Act (ARRA). The rest appear to reflect changes in categorization rather than substantial innovations.

The wide popularizations of two first-appearance features, Build America Bonds (BABs) and the make-whole call provisions, were driven by a temporary federal stimulus program. The ARRA also created Qualified School Construction Bonds (QSCB) and Recovery Zone Bonds (RZB). BABs and RZBs appear in 2009, peak in 2010, and end abruptly and completely as their programs expire. QSCBs continue as a small program after 2011 (less than a 0.25 percent market share). The diffusion and use of these debt types were directed by the incentives designed into the program. Analysis of the BABs program can be found in Luby (2012), Cestau et al. (2013) and Liu and Denison (2014). The make-whole call feature first appeared in 1997. It was used occasionally until 2008, exploded in usage with the BAB program, and then remained fairly common thereafter. The BAB program also raised the designation of bonds as federally taxable from a rare to a common feature, but taxable bonds pre-dated the study period.

Among the rarely used first-appearance innovations, the only other one that appears substantial is the naming of an investor relations agent. In total, 11 of the features are new debt types. These include Combined Anticipation Notes, Tax and Revenue Anticipation Notes and Federal Aid Anticipation Notes. It seems likely that these notes are not fundamentally different from securities in some other categories. For example, while there are no Federal Aid Anticipation Notes in 1992 or 1993, there are Grant Anticipation Notes and Appropriation Anticipation Notes. Likewise, there are no Tax and Revenue Anticipation Notes in the data before 1995, but there are dozens of Tax Anticipation Notes and Revenue Anticipation Notes before 1995 tracked separately.

4.2 Popularization

The literature does not provide a standard market share threshold for designating an innovation to be popularized, so I will discuss a few different levels. There are 18 features that were consistently above a 10 percent market share in the earliest years of the data. These cannot be considered innovations during the study period as they were already well established. In contrast to these common features, there are 97 features that can be designated “Never Popularized.” These include 47 features that were never observed with a market share above 1 percent, 31 that peaked between 1 and 3 percent, and 19 features that had peak shares between 3 and 5 percent[9]. In total, 13 features exhibit medium growth from market shares below 5 percent to market shares between 5 and 12 percent. In total, 17 features began with market shares below 10 percent and achieved high growth, expanding their market share by over 10 percentage points.

Before moving on from the features that individually never had a market share above 5 percent, it is interesting to consider their use in aggregate. Figure 1 shows the percent of par value that was issued with one, two, or three or more uncommon features in each year. The first thing we learn from this graph is that it is very common for securities to have a few uncommon features[10]. On average, 23 percent of the par value is associated with one of the uncommon features and another 5 percent is associated with two uncommon features. This suggests that the market is not dominated by a handful of plain-vanilla securities. Instead, it appears that market participants are experimenting with dozens of uncommon features, and willingness to issue a security with an unusual feature is widespread. This should provide fertile ground for innovation. Figure 1 also displays a time trend. Use of uncommon features that remained uncommon rose from 1998 until 2005. This apparent experimentation has declined from over 40 percent of the market to below 25 percent since the financial crisis.

Among the 14 “Medium Growth” features that began with market shares below 10 percent and had a medium amount of growth, 5–10 percentage points, 8 involved the naming of uncommon agents or more than the standard number of agents. These included the naming of registration, fiscal, transaction, auction and exchange agents, as well as additional trustees and an unusually high numbers of underwriters. There were increases in the share of issuance that was designated alternative minimum tax qualified and bank qualified.

Finally, there are 17 “High Growth” features which display a strong innovation pattern. All of these began with market shares below 10 percent in 1992 and 1993, and then grew their market share by between 10 and 66 percentage points. The highest growth was observed for premium bonds which rose from a 5 percent market share in 1993 to a 71 percent market share in 2014. This rise mirrored the decline of discount bonds. If bonds with an offering price different from their face value are considered together, they were already above a 50 percent market share at the beginning of the study period. High growth was also associated with several of the agents including lead underwriters (two or more), financial advisors (two or more), remarketing agents, tender agents and escrow agents. Taxable bonds were in use from the beginning of the study period, and their market share spiked and fell with the taxable BABs. Several of the other high-growth features appear to be administrative in nature, such as denominations greater than $5,000 and non-standard interest calculation frequencies. The four features which will be examined in greater detail in Section 5 are all in the “High Growth” category.

4.3 Persistence

If one strictly applies a persistence criterion, almost all the potential innovations would be disqualified. Among the 18 features that made a first appearance after 1993, 13 fell all the way back to zero by 2014 while the rest maintain market shares below 5 percent. The most persistent first-appearance security was tax and revenue anticipation notes which continued to have a market share between 3 and 6 percent in later years. The make-whole call provision remained near a 3 percent market share. Among the 13 “Medium Growth” (5–10 percentage points) features, only three maintained market shares above 5 percent in the last three years of observations. These were the naming of multiple trustees, multiple registration agents, or a fiscal agent. Among the “High Growth” innovations that started below 10 percent and grew by more than 10 percent, most exhibited a rapid decline following the financial crisis. Of the 17 features that were popularized after 1993, 12 features’ market shares fell below 10 percent but remained above zero. Only five maintained market shares above 10 percent through 2014. Those were the use of more than one financial advisor, naming more than one lead underwriter, non-standard call notices, premium bonds and the use of bond proceeds for primary and secondary education[11]. Requiring persistence at a high (>10 percent) level of market share would limit the potential innovations to a very short list of mostly inconsequential features. The innovations selected for further analysis display a moderate level of persistence with market shares in the last years of the data between 4 and 10 percent.

4.4 Selecting features for further analysis

To understand the diffusion and adoption processes, the best features are those that experienced a large increase in their market share during the study period. Features that were already widely in use, and those that remained uncommon or rare would not be as informative. The 20 features designated “High growth” or “First Appearance, Popularized” in Table I are the best candidates. From these, I have selected four to analyze as examples of adoption and diffusion: variable rate securities (VRS), put options, corporate backers and derivatives[12]. In the next paragraphs, I will provide a brief description of each item. The descriptions will provide background for thinking about why and when they were popularized. I will also give some details about how each of these items is tracked in the Mergent data.

Several of the “High growth” and “First Appearance, Popularized” innovations were not selected for specific reasons. Three expanded due to a unique federal stimulus. While interesting, the experience of these features cannot be generalized to less proscribed innovations. Two of the high-growth features appear to be ancillary to the use of variable rates. Over 97 percent of issues that name a remarketing or tender agent are VRS, so analyzing the use of these agents would add little information if the use of variable rates is analyzed. Some of the high-growth features appear to be administrative, such as naming an escrow agent, or unremarkable, such as the use of bond proceeds for primary and secondary education. To keep this analysis tractable, some potentially interesting innovations must be left to future work including the use of special mandatory calls. The other “High Growth” features not subject to further analysis here are non-standard call notice requirements, denominations above $5,000, non-standard interest calculation frequencies, non-standard interest rate calculation formulas, more than one financial advisor, more than one lead underwriter and premium bonds.

4.4.1 Variable rate securities

VRS were embraced by issuers because they appeared to reduce interest costs while maintaining a steady, predictable series of payments. Detailed descriptions of the use of VRS can be found in Stewart and Cox (2008), Stewart and Smith (2012), Singla and Luby (2014) and Luby and Kravchuk (2013). These bonds were appealing to investors because they reduced interest rate risk for the investors. The interest rates on the VRSs could be linked to an index, such as the London Inter-Bank Overnight Rate, or it could be reset periodically via auctions arranged by underwriters or brokers. The interest rate risk is borne by the VRS issuers, and in most cases, the issuers sought to hedge that risk by entering swap contracts. The issuers paid a constant, and thus predictable, stream of payments to the holder of the swap, and that investor paid additional interest to the variable rate security holder when a rate reset required it.

In instances in which not enough outside investors bid on the ARS, the underwriters and brokers could prevent the auction from failing by bidding themselves. As the financial crisis unfolded in 2008 and 2009, several forces combined to swiftly curtail the use of VRS, ARS and their associated swaps. Interest rates fell to historic lows, making locking in low fixed rates much more attractive relative to variable rates. For ARS, the underwriters and brokers who had provided liquidity by bidding in the auctions stopped doing so. Before 2007, auctions almost never failed. During the financial crisis, the auctions began to fail frequently, and this let the securities default to the contractually predetermined maximum interest rates. Most ARS carried bond insurance, which enhanced their liquidity. Several of the major insurers experienced their own crises and credit ratings downgrades that reduced or eliminated any advantage they could provide for ARS issuers. Falling interest rates caused the VRS- and ARS-related swaps to have a negative value, and issuers were obligated to post collateral. Finally, issuers who wanted to exit their swaps, because they could now refinance the hedged debt to lower fixed rates, had to pay termination payments.

The financial crisis revealed that, while interest rate hedging protected issuers from one risk, the need for banks to provide liquidity exposed the issuers to another, unrecognized risk. The lower costs that the issuers enjoyed in the years before the crisis were a result, in part, of not properly insuring against this risk. If rates rise, and issuers again seek to access long-term borrowing at lower rates, we will see if the market offers some type of insurance against auction failures. Now that this risk is understood, issuers may decide ARS are unfavorable if this risk must also be hedged at a cost.

The variable rate innovation is represented in the Mergent “coupon code” variable by the labels “variable,” “adjustable,” “floating auction rate,” “floating,” “flexible rate,” “floating rate at floor,” “inverted,” “index linked” and “linked inverse floater.” Of these, the “variable” designation appears with 66 percent of the VRS par value, and the “floating auction rate” designation appears with 18 percent of the VRS par value. Floating ARS by themselves never rose above 10 percent of the total market par value. From 2002 to 2007, they were between 5.0 and 7.8 percent.

4.4.2 Put options

The story of the put option is similar to that of the VRS. If interest rates rise, bondholders will want to put their bond back to the issuer in order to recover and reinvest their funds (Temel, 2011). The issuer of a puttable bond bears the risk of having to return the full principal on several dates before the maturity date. Issuers recognize that they cannot bear this risk directly, so they engage a bank or other liquidity provider. When interest rates are very low (having nowhere to go but up), liquidity provision is expensive because the probability that the bonds will be put back is higher. We see a marked decline in put options after the financial crisis, as issuers decided against offering them.

Put options are tracked by a dedicated variable that takes the values of “Y” or “N.” The “put option” variable is populated for more than 99 percent of observations in every year of the data, so it is likely that the increase in usage observed was a real expansion rather than a mere shift from not recording the option to recording it.

4.4.3 Corporate backers

The partnering of municipal issuers with corporate backers is a long-standing practice that experienced a sharp increase during the study period. Corporate entities, including hospitals and postsecondary schools, are expected to provide the funds to repay the principal and interest on securities issued through the partnerships (Cavallaro, 2012; Mincke, 2012). These private partners default and fail much more frequently than municipal corporations. The municipal partners would be responsible for repaying outstanding bonds, but they generally purchase insurance to cover this risk.

Partnership with a corporate backer is also tracked by a dedicated variable that is populated in all years of the data. The variable lists a code corresponding to a specific corporate entity that can be cross-referenced in a separately maintained Mergent directory. For the analysis below, the presence of any code in this variable is converted into an indicator that the feature was used.

4.4.4 Derivatives

Derivatives differ from other municipal securities in that they are created in the secondary market. An investment bank or other financial service provider can purchase a municipal security and then sell the right to receive part of the coupon payment (Mysak, 2012). The derived portion of the payment can be a fixed or variable portion, and the firm can retain or sell the remaining portion of the payment. These derivative securities enable investors to take long or short positions on various indices (Felstein et al., 1995).

It is important to note that the municipal issuers do not choose whether their bonds become the basis of a derivative product. The debt has already been issued, and the terms of the bond set. Indirectly, derivatives may increase demand for certain types of municipal securities, and issuers may find it advantageous to offer derivative-friendly terms. In the Mergent data, the underlying bond issuer is recorded in the “issuer” field, but the interpretation of issuers’ trends is different. For the other innovations, we are observing which issuers choose to issue the innovative product. With derivatives, we are observing which issuers the investment bankers choose as the basis of their derived security. While the investors may be seeking exposure to certain types of interest rate risk, they may be selecting municipal bonds because of the bonds’ very low default and liquidity risk. Also, we must keep in mind that derivatives are not subject to the same disclosure requirements as bonds or notes. Some derivatives are created as contracts and are not assigned CUSIPs to trade as securities. To the best of my knowledge, comprehensive estimates of municipal-based derivative volumes are not available. The estimates presented here based on the Mergent data can be regarded as a lower bound. The volume of all derivatives activity is estimated by the Bank of International Settlements using survey data, but derivatives related to municipal securities are not tracked separately (Wooldridge, 2016). While several studies have been published on the use of derivatives by municipal borrowers, the use of municipal securities to create derivatives has received little research attention (Luby, 2012; Luby and Kravchuk, 2013; Singla and Luby, 2014).

“Derivative” is one of 29 designations tracked in Mergent’s “debt type” variable. Like the put option, hundreds of issues are designated as “derivatives” even in the first years of the data, so it was something being recorded throughout the study period. Out of the 29 debt types, 9 were added after 1994, a circumstance which demonstrates that Mergent does expand this category when a differentiated product appears[13].

When market share calculations are presented below, the derivative par values are always excluded from the denominator to avoid duplicating dollars counted when the bond was originally issued. The time series and shares for derivatives represent how large the volume of derivatives was relative to the volume of non-derivative securities issued for the particular year and level of aggregation (issuer, state, total market, etc.).

Recalling the definition of innovation discussed in Section 2, variable rates, put options and derivatives would be considered innovations by Miller’s (1986) and Finnerty and Emery’s (2002) definitions. Variable rate bonds did lower issuers’ costs for many years. However, they did not thrive after the changes in interest rates and regulation, so they could be considered “not significant” if they do not return to prominence at some future date[14]. In Tufano (2003), innovations have to make a first appearance in the data and then become popular. Variable rates, derivatives, put options and corporate backers existed before 1992, so they could be called innovations of some earlier time. By the second part of Tufano’s definition, which requires popularization, all four features are innovations because they were all extensively popularized used with more than 10 percent of the issuance volume at some point during the study period.

5. Adopters and diffusion of the innovations

5.1 Evolution of the innovations’ market share

Figure 2 presents the market share of the four innovations during the study period. In the early and mid-1990s, variable rates were used with 10 percent of the market share or less. From 1998 through 2007, the market share of variable rate bonds was on an upward long-term trend. The market share of bonds with put options followed a similar path, including a spike in 2007. The market share of corporate-backed securities and the relative volume of derivatives also grew most rapidly between 2004 and 2008. Use of all four types of innovations plummeted in 2009. Derivative and corporate-backed securities remained above pre-2004 levels but below their peaks during the recovery.

In Figure 3, the market shares are presented as a percent of the bonds issued rather than as a percent of the par value. The first thing we notice is that these unweighted percentages are much lower, with all remaining below 8 percent. Another disconnect is that the rise in the percent of issues with variable rates, derivatives, and corporate backers was much more concentrated in the period 2004–2008 than was the rise in their market share.

The four innovations of interest can often be found used in combination. Figure 4 illustrates which of these combinations are most common, and it provides nominal dollar totals. In the 1990s, the innovations were in use, but they were associated with less than $30bn of issuance. The usage climbed rapidly to a peak of $278bn in 2008. Most of the innovation-linked issuance is associated with 5 of the 15 possible combinations of the features. The largest type by volume is variable rate securities with put options, followed by corporate-backed securities. The other three major types are those with variable rates only, variable rate derivatives, and variable rates with a put option and a corporate backer. All the other types are found with relatively small par values.

5.2 Yields and use of proceeds from securities with innovative features

As discussed above, a variable rate bond transfers the interest rate risk from the investor to the issuer. With that transfer comes a shift in which party would desire a variable rate. When rates are expected to rise, investors would prefer a variable rate security, and issuers would prefer to lock in the relatively low current rate. When rates are expected to fall, investors would prefer a fixed rate security, and issuers would prefer the variable rate. In Figure 5, the initial yields observed in the data are plotted. Over the study period, they have drifted downward. The peak market share of variable rate securities followed a small rise in yields for all other municipal securities. Both issuers and investors have the ability to insure against their interest rate risk. In the mid-2000s, rates were low relative to those during the previous 20 years, so investors could reasonably expect that they would rise, making a variable rate security preferable. It is less plausible that issuers foresaw the decline in rates following the recession, so they should have used some of the savings released by issuing variable rate debt to purchase offsetting insurance against rate hikes.

Table II displays the top five most common uses of bond proceeds associated with the innovative features. VRS with and without put options are most often linked to general purpose or public improvement projects. Among the top five most common uses of innovative securities, most of the other positions are held by health care, higher education and housing projects. For comparison, the percentages are displayed for securities without innovations. General purpose/public improvement uses are by far the most common for non-innovative securities, followed by primary/secondary education. Almost half of all derivatives are created using general purpose municipal bonds.

Of the two major types of municipal debt, general obligation and revenue bonds, the innovations were more common among revenue bonds. Figure 6 shows that use of variable rates, put options and corporate backers by general obligation issuers was only slightly elevated in the 2000s[15]. Among revenue-backed securities, the rise from an approximately 5 percent market share to a 35 percent market share was a nearly linear trend over a 16-year period.

5.3 Adopters

Leveraging the individual issuer observations in the Mergent data, it is possible to explore which types of issuers were the adopters of the innovations. In Figure 7, the market share of securities with variable rates, put options or corporate backers is plotted for five types of jurisdictions. In most years, county governments were issuing with the greatest share of their volume having the innovative features. Special districts, cities and school districts appear to increase their usage of the features by 5–10 percent of par values around the year 2000. Between 2000 and 2010, the innovation market shares of states, special districts and counties were similar in both their levels and trends. All five series peak in 2008 before falling sharply in 2009. It is notable that use of these innovations does not seem to be interacting with the qualitative differences between the types of issuers. States are sovereign borrowers without a clear recourse to bankruptcy while the other issuers are not sovereigns and can file for protection from creditors in federal court. Yet state use of the innovations is very similar to that of counties and special districts. The division of local government services between counties, cities and special districts is both highly varied and fluid. Any function performed by counties in some regions is performed by cities in other regions and vice versa, yet the use of innovative municipal securities by counties is 5–10 percentage points higher in most years both before and after the financial crisis. The market shares for school districts remains fairly low throughout the study period.

The values in Figure 7 were calculated with each issuer having equal weighting. However, the municipal securities market is highly skewed, with a few large issuers of each type issuing the majority of debt outstanding. Approximately 40,000 state and local governments had debt outstanding sometime between 1992 and 2015. When the issuers are categorized so that each category corresponds to one quarter of the debt outstanding in the year, the third and fourth quartiles contain approximately 32 states and 19 substate governments with the greatest debt outstanding. The second quartile contains 16 smaller states and approximately 425 midsized local governments. The lowest quartile contains the remaining 39,000 local governments. If we group issuers by their debt outstanding rather than their type, we can see in Figure 8 that the issuers backing large amounts of debt were similar to smaller issuers until 2000. After 2000, the top three quartiles diverge from the bottom quartile in the extent of their use of the innovative securities[16].

5.4 Within-state diffusion

The three features of interest that issuers can select – variable rates, put options and corporate-backed securities – were all uncommon but existent during the 1990s. We would like to know if the expansion of their use was via issuers in more states using the feature, issuers in the same states using the feature with greater volumes (an intensive or extensive expansion) or both. Did early adopters expand their reliance on the innovations, or did late adopters catch up after observing their peers? In Table III, the market share by state between 2000 and 2008 is differenced from the market share during from the 1990s. There is a strong positive correlation between the extent of use by issuers within a state in the earlier and later periods. All states grew their market share of securities with the innovative features. Vermont, Wyoming, West Virginia and Montana issuers used the innovative features with a high portion of their issuance in both periods. Interestingly, the largest states by population were near the middle of the distribution on the market share measures. Also, being home to the nation’s financial centers did not cause New York, California or Illinois to be early or aggressive adopters of these financial innovations.

5.5 Fitted models of adoption of innovations

To supplement the information that can be gained from the figures, this section will present the results of several fitted models. The first models consider the relationship between yields and the use of innovative features. The next two sets of models have dependent variables that are either indicators of an issuer’s using an innovative product sometime during the year or the share of the issuer’s par value within a year that is associated with the innovation. The independent variables include the type of issuer, the quartiles of revenue and debt outstanding, budget growth between COG years and state and year fixed effects[17]. I also include lagged measures of usage of the innovative product aggregated to the state level. The lagged state aggregate measures give some insight into whether the innovations were diffusing through issuers in the same state. Bond issuers might learn about an innovative product from similar issuers in their state through neighboring jurisdictions’ disclosures, state regulations, professional organizations or shared regional financial service providers. Finally, I present models that estimate time trends of adoption and test whether they differ by the type of issuer.

5.5.1 Yield models

If seeking lower borrowing costs is the primary motivation for adopting variable rates and put options, we would expect to see these features more frequently when yields are higher. Likewise, corporations and non-profits have more to gain by partnering with municipal governments when interest rates are relatively high. However, the selection of these features based on their pattern of innovation makes the relationship less obvious. During the study period, the highest interest rates prevailed in the first eight years, when the innovative features were uncommon. In the middle years, from 2000 to 2008, interest rates were lower than they had been in the 1990s, and the innovative features experienced rapid growth. Since the financial crisis, both yields and use of the innovative products, has been low. When we estimate the relationship between yields and innovative features, we arrive at counterintuitive coefficients that are small in magnitude or even negative in some cases.

Table IV presents regression estimates of the relationships between the prevailing yields, the spreads of the innovative securities, and the probability that a security will have the feature. The yield measure in the models for any innovation, variable rates, put options and derivatives are the three month moving average of the offering yields of all municipal securities that do not have any of the innovative features. The yield measure is meant to capture the cost of borrowing that an issuer faced while the spread reflects the advantage the innovative security might have realized for them. Both the yields and spreads are lagged by one month so that no security can contribute to both the dependent and independent variables. In the corporate-backed model, the yield variable is the Bank of America Merrill Lynch US Bond Yields Broad Market Index (yield to worst). The regressions include an issuer fixed effect, so the coefficients are identified with the variation in yields and spreads between issuances by repeat issuers. The observations are weighted by par value. Derivative securities observations are only included in the last regression, where their dependent variable value is 1 and all other securities have a value of 0.

A standard deviation of the yields is close to 1 for municipal bonds and close to 1.5 for the corporate bonds. The probability of a bond using a variable rate is 1 percentage point higher when yields are higher by one standard deviation. Use of put options responds similarly with a coefficient of 0.008. Over the study period, variable rate bonds display a negative relationship between their spread and their probability of usage. In Figure 5, we can see that the yields on variable rate securities and all other municipal securities converged during 2006 and 2007, just as the use of VRS approached its peak. In contrast, the spread on put option securities remained similar to what it had been since 2000. The coefficient on the put option spread measure suggests a 1 percentage-point difference in the spread would correspond to a 1.3 percentage-point difference in the probability a put option is used. A difference of this size is approximately 11 percent of the sample mean of the indicator of having a put option. The use of corporate-backed municipal bonds grew during a period of falling yields, and this causes the overall relationship to be negative.

5.5.2 Characteristic models

Table V provides descriptive statistics of the dependent and independent variables used in the regressions presented in Tables VI and VII. The first set of coefficients in Table VI relates the probability that an issuer issues a security with the innovative features in a year. The observations are issuer-years for all state and local governments reported in the COG from 1995 to 2015. Observations for 1992 through 1994 must be excluded because the lagged values are not available for those years. In Table VII, the observations are only governments that issued within a year because the dependent variable, share of par with the innovative feature, can only be calculated if some issuance is observed. Cities and towns are the most numerous type of government, and they will be the omitted type. Note that states represent less than 1 percent of potential issuers and observed issuers. The quartiles of debt were selected so that issuers in each quartile back one-quarter of all debt outstanding. With the lowest-revenue and zero-debt governments serving as the omitted category, the quartile indicators identify the large and very large participants in the municipal securities markets. Finally, the change in annual expenditures is calculated between the waves of the COG data and assigned to each intervening year. This is meant to represent whether the governments are rapidly expanding or contracting in real terms. The governments with fairly stable budgets are the omitted category.

The final sections of Table V describe state aggregate measures of use of the innovative features. For the dichotomous issuance model, we expect that an issuer is more likely to issue an innovative security if more issuers in their state have been using the feature in the recent past. The count of issuers is scaled by the state’s population. This measure will be dominated by small issuers, and the state government can add only one to the count. In contrast, state governments can be a major driver of the state’s aggregate market share measure because states usually issue a major portion of all new bonds that are aggregated into a state-year observation.

In Table VI, we can see that, conditional on the other measures, state governments are far more likely than cities (the omitted category) to issue variable rate, put option and corporate-backed securities. State bonds are more likely than city bonds to be used in derivatives. The differences between cities, counties, special districts and school districts are small in magnitude even when they are significant statistically. Using innovative products or backing a derivative is strongly positively related to high levels of debt outstanding. Budget growth and decline are positively and negatively correlated with use of innovative products, but the magnitudes are small. This suggests innovations are not being disproportionately sought out by either booming or struggling municipalities. Finally, we do find evidence that having more issuers in one’s state that used innovations in recent years makes an issuer more likely to use the innovation in a subsequent year. For example, a one standard deviation (0.194) increase in issuers per 100,000 residents in a state in the previous year would correspond to an increase of 0.13 percentage points in the probability that an issuer uses a variable rate security in the next year.

A parallel set of models is reported in Table VII that explores the relationship between the characteristics of the issuers and the share of their par value that has the innovative features. States have shares that are 9.6 percentage points higher than that of cities for variable rates. State’s market shares with put options is 5.8 percentage points higher than that of cities, and the equivalent figure for corporate-backed securities is 6.8 percentage points. The innovation market shares are also higher for counties and special districts, in the range of 17–28 percent of a standard deviation. Conditional on the other independent variables, issuers with higher revenues do not have consistently higher shares of their issuance associated with the innovations. Issuers with low but positive amounts of long-term debt outstanding are less likely to issue debt with innovative features relative to the reference category, which includes governments with no debt outstanding. Issuers in the second quartile of debt outstanding use significantly higher shares of variable rate and put option securities. The coefficients on the indicators of being in the third and fourth quartile of debt outstanding are positive but not significant in their relationship with using variable rates, put options, corporate backers and combinations of these. A clear positive relationship is present between the level of debt outstanding and the ratio of derivative issuance to securities issuance for an issuer. The coefficients on the budget growth measures in the market share models are very small and insignificant in most cases.

The lagged aggregate state market shares are predictive of issuers’ market shares for each type of innovation, but the magnitudes are modest. For example, a one standard-deviation increase in the one-year lagged state aggregate market share of securities with put options would correspond to a 4.4 percent of a standard deviation increase in market share for the average issuer ((6.798×0.106)/16.238=0.044).

5.5.3 Time trend models

In the time series graphs (Figures 7 and 8), we saw the paths of the mean market share by type of government and quartile of debt outstanding. These lines trace midpoints of distributions with wide variances. We need a model estimate to test if the pace of adoption was significantly different between governments of different types and debt levels.

Table VIII presents three sets of results from models that focus in on the period of rapid expansion of the innovations between 2000 and the peak year of 2007 or 2008. These models give a slope coefficient for the adoption of innovations and enable us to determine if the differences are significant. To this end, the models include a constant, a demeaned linear year measure, an indicator for the type of government or quartile, and interactions of each of these. With the interactions included, the coefficient on the year can be interpreted as the pace of adoption (percentage point increase per year) by the issuers in the omitted category. Adding the coefficients from the interaction of the year and the type indicators to the coefficient on year provides an estimate of the slope of adoption by the other groups of issuers.

Table VIII displays trends in the overall use of securities with innovations that are negative for cities, counties and the issuers with the least debt outstanding. In contrast, states were increasing their use of innovative securities by 2.06 percentage points per year (−0.20+2.26). Special districts were raising their market share by 0.51 per year (−0.20+0.71). Cities were slowly expanding their use of corporate-backed securities while reducing their use of VRS and put options. States were more aggressively expanding their use of VRS at a rate of 1.4 percentage points per year (−0.30+1.70). States were also expanding their use of corporate-backed securities at 1.86 percentage points per year (0.33+1.53).

When the issuers are grouped by their debt outstanding, the lowest-quartile issuers were decreasing their market share of variable rate and put option securities by 0.39 percentage points and 0.17 percentage points each year. Issuers in the top three quartiles by debt outstanding, the approximately 600 largest issuers, were all increasing their use of the innovative securities by 0.23 percentage points to 2.25 percentage points per year. The last set of results shows issuers grouped by their issuance in the last three years. With this division, the third- and fourth-quartile issuers appear to be increasing their use of the innovations at a pace of 0.57 percentage points to 2.02 percentage points each year. The first quartile comprises issuers who have issued no debt in the last three years. Each year, these infrequent issuers are more likely to appear with corporate-backed debt.

Regarding the use of an issuer’s debt for derivatives, all models report positive time trends for all types of issuers. However, issuers with the most debt outstanding and with the highest recent issuance volume are associated with the most rapid growth of use as a derivative underlying. The growth rates for cities and first-quartile issuers are positive and significant, but close to zero in magnitude. Growth rates for states, third-quartile issuers, and fourth-quartile issuers (by quartile of debt outstanding or recently issued) are 1 percentage point to 2 percentage points per year.

6. Conclusion

The Mergent data appear to be recording the vast majority or possibly all of the municipal securities issued each year. Based on the appearance of new values, categories and flags in later years, it seems that Mergent does adapt its recording to reflect innovations. If there is an innovation not shown in the data, it would have to be something that Mergent cannot observe or record (such as novel bank loan arrangements) or something that its client base is willing to let it omit from the data.

The Mergent data do reflect the first-time arrival of some new securities and features, but it appears that very few of these innovations have become popular since 1992. Those that have, such as BABs, were created by legislation rather than market participants. Instead, innovation in the municipal securities market has taken the form of rare securities and features’ becoming widespread, as was the case for variable rate securities, put options and derivatives. The persistence of these innovations is an open question. Their market shares have fallen dramatically as we entered an era of historically low interest rates. Whether the innovations return will not be known until rates rise substantially from current levels.

It is also worth noting that, except for the ARRA-related securities, no new innovations appear to have emerged in response to the crisis. Despite dramatic changes in the market, such as the downgrading of the monoline bond insurers, we have not seen new products or features appear in response. We have not even seen a product rise significantly in market share by continuing to be issued at steady levels while aggregate issuance fell. This analysis allows one additional characterization of innovations in municipal securities markets: they appear to arise during times of stability and expansion rather than during times of distress or recovery. We should keep in mind the possibility that innovative intermediation has migrated to markets not covered by this data (Bergstresser and Orr, 2014). For example, banks may be providing municipalities with loans that have features like those observed to decline here. The loans could also have entirely different innovative features. This would not change the conclusions about the municipal securities markets specifically, but it could have implications for which investors are bearing the risk and realizing the returns from municipal debt.

Descriptive and regression analyses show that innovations are adopted faster and with a greater share of issuance by the approximately 600 largest issuers that back 75 percent of the debt outstanding. Smaller issuers also use innovative products, but to a lesser extent. There is evidence that use of innovations does diffuse through states. Conditional on year fixed effects and other observables, issuers are more likely to use an innovative security or feature if it was used more frequently or with a greater share of issuance in their state in the previous two years. A relationship can be detected between budget growth and use of innovations, but it is small in size.

Financial innovations present an opportunity as well as a risk. If we observe that primarily the largest, most sophisticated debt issuers are using new products, we may ask if smaller, less sophisticated issuers could also benefit from these products. Less sophisticated issuers may not be aware of the innovations, or they may avoid them because they are unable to assess the risks. In this case, the technical assistance of sophisticated, disinterested intermediaries, such as state bond banks, may be able to overcome the information asymmetries, realize lower borrowing costs for taxpayers, and create new investment opportunities for bondholders. It is possible that fixed costs make the innovations only cost effective for a large issuance or refinancing. That would also favor aggregation of debt issuance. On the other hand, the recent experience with innovative products did not end well for many municipal issuers (Stewart and Smith, 2012; Denison and Gibson, 2013). If more municipal issuers had engaged the innovations of the last cycle, they, too, would have experienced failing rate auctions, collateral calls and swap termination fees.

Figures

Market share of securities with features never widely adopted

Figure 1

Market share of securities with features never widely adopted

Financial innovations market share

Figure 2

Financial innovations market share

Financial innovations’ share of issues

Figure 3

Financial innovations’ share of issues

Par value issued with innovative features

Figure 4

Par value issued with innovative features

Initial yields by security type

Figure 5

Initial yields by security type

Issuer-selected innovation market share by security type

Figure 6

Issuer-selected innovation market share by security type

Issuer-selected innovation market share by type of issuer

Figure 7

Issuer-selected innovation market share by type of issuer

Issuer-selected innovation market share by quartile of debt outstanding

Figure 8

Issuer-selected innovation market share by quartile of debt outstanding

Designation of features by innovation criteria

Market share
Initial Growth Peak Designation Count
10.2 to 96.6 0 to 20.8 10.9 to 96.6 Already popularized 18
<0.1 to 3.8 0 to 3.8 <0.1 to 4.9 Never popularized 97
0 <0.1 to 1.7 <0.1 to 1.7 First appearance, not popularized 15
0 6.3 to 23.4 6.3 to 23.4 First appearance, popularized 3
0.6 to 8.3 0 to 4.9 5.0 to 10.4 Low growth 14
0.1 to 3.2 5.6 to 9.5 5.7 to 12.0 Medium growth 13
0.7 to 5.5 11.1 to 65.6 15.4 to 70.8 High growth 17
Total features 177

Notes: Data are from the Mergent Municipal Bond Securities Database. “Initial” is the higher of the market shares in 1992 and 1993. “Peak” is the highest annual market share observed for the feature from 1992 to 2014. “Growth” is the difference between the feature’s initial and peak values. If the feature’s peak market share is in 1992 or 1993, then the growth is zero. The values for derivatives are the ratio of the derivative par value to the total market issuance rather than a standard par value market share. The pattern of this ratio places derivatives in the “High Growth” category

Most common uses of bond proceeds by innovation type

No innovation Variable with put option
General purpose/public improvement 34.6 General purpose/public improvement 16.5
Primary/secondary education 14.3 Water and sewer 8.3
Water and sewer 8.4 Student loans 7.5
Higher education 6.4 Higher education 7.3
Electric utilities 4.4 Hospitals 5.6
Corporate Variable
Hospitals 32.5 General purpose/public improvement 20.7
Higher education 21.2 Student loans 11.1
Other health care 13.1 Multi-family housing 8.5
Pollution control 3.5 Higher education 8.5
General purpose/public improvement 3.1 Single family housing 5.9
Variable, put option, corporate Variable derivatives
Hospitals 28.4 General purpose/public improvement 47.5
Other health care 13.7 Water and sewer 6.6
Higher education 11.4 Primary/secondary education 5.1
Multi-family housing 6.9 Higher education 4.3
Pollution control 6.8 Other health care 3.8

Note: Data are from the Census of Governments and the Mergent Municipal Bond Securities Database

Issuer-selected innovation market share by decade

State 1992–1999 2000–2008 Difference State 1992–1999 2000–2008 Difference
VT 27.92 57.77 29.84 MI 16.94 29.55 12.61
WY 36.83 51.25 14.42 ID 9.75 28.44 18.69
WV 19.60 49.61 30.01 AZ 10.32 28.39 18.07
MT 26.12 44.95 18.83 OH 18.53 28.17 9.64
MS 17.49 44.17 26.68 MA 13.85 27.87 14.02
NH 24.72 43.53 18.81 ND 17.81 27.31 9.50
LA 25.20 43.33 18.13 NY 13.40 26.91 13.51
NC 16.00 42.76 26.77 VA 13.24 26.37 13.13
PA 15.59 41.16 25.57 AR 17.56 26.26 8.70
MO 21.91 40.08 18.17 CA 13.42 26.24 12.83
KY 14.66 39.21 24.56 WI 11.21 24.88 13.67
TN 25.82 38.47 12.65 ME 8.61 23.89 15.28
IN 22.09 38.04 15.95 SC 18.29 23.87 5.58
OK 17.60 37.44 19.84 RI 18.30 23.51 5.21
MN 12.07 37.30 25.23 SD 9.69 23.48 13.79
CO 15.26 36.94 21.68 CT 17.10 23.29 6.19
UT 17.37 35.38 18.00 NV 11.19 23.27 12.09
GA 22.30 34.37 12.07 TX 12.37 21.36 8.99
AK 10.08 33.81 23.73 NJ 12.54 19.16 6.62
AL 22.04 33.80 11.77 WA 8.29 17.98 9.69
IA 19.42 31.74 12.32 DE 4.10 17.21 13.11
NM 15.99 31.17 15.19 KS 14.06 15.99 1.93
FL 17.50 30.99 13.49 OR 7.54 15.37 7.82
MD 14.39 30.46 16.07 HI 6.05 14.44 8.39
IL 17.42 29.77 12.35 NE 2.00 14.06 12.06

Notes: Data are from the Mergent Municipal Bond Securities Database. Innovations included are variable rates, put options and corporate backers

Yields and use of innovations

Any (Var/Put/Corp) Variable rates Put option Corporate Derivative
Descriptive statistics
Indicator of innovative feature (weighted by par value)
 Mean 0.218 0.144 0.116 0.105 0.027
 SD 0.429 0.371 0.341 0.321 0.181
Yields on municipal bonds without the innovative features or corporate bonds
 Mean 4.134 4.134 4.134 4.302 4.136
 SD 1.075 1.075 1.075 1.532 1.065
Spread between yields on bonds with and without the innovative features
 Mean −0.145 −0.648 −0.847 0.452 0.131
 SD 0.769 1.204 0.738 1.305 1.701
Regression estimates
Yields, no innovation 0.015*** 0.010*** 0.008*** −0.037*** −0.003***
(0.000) (0.000) (0.000) (0.000) (0.000)
Yield spread, innovation 0.062*** −0.007*** 0.013*** −0.047*** −0.005***
(0.000) (0.000) (0.000) (0.000) (0.000)
Issuer fixed effects Y Y Y Y Y
Constant 0.166*** 0.100*** 0.092*** 0.293*** 0.038***
(0.001) (0.001) (0.001) (0.001) (0.000)
R2 0.263 0.179 0.165 0.264 0.048
n 2,818,187 2,818,187 2,818,187 2,392,636 2,838,242

Notes: Data are from the Mergent Municipal Bond Securities Database and the Bank of America Merrill Lynch US Bond Yields Broad Market Index. Yields and spreads are averages over the three months preceding the security’s month of sale. The dependent variable is an indicator of the security utilizing the innovation. *p<0.05; **p<0.01; ***p<0.001

Descriptive statistics

All state and local governments Municipal securities issuers
Mean SD Mean SD
Indicator of use Percent of par value
Any (Var/Put/Corp) 0.006 0.079 6.979 23.719
Variable rate 0.004 0.067 4.418 18.776
Put option 0.004 0.061 3.364 16.238
Corporate 0.004 0.061 3.787 17.727
Derivative 0.002 0.044 0.446 4.659
City/town 0.367 0.482 0.403 0.491
State 0.000 0.022 0.008 0.089
County 0.031 0.173 0.122 0.327
Special district 0.353 0.478 0.115 0.319
School district 0.140 0.347 0.319 0.466
Type missing 0.108 0.311 0.033 0.179
Revenue – 1st quartile 0.201 0.401 0.007 0.085
Revenue – 2nd quartile 0.201 0.400 0.021 0.145
Revenue – 3rd quartile 0.201 0.401 0.130 0.337
Revenue – 4th quartile 0.201 0.401 0.731 0.443
Revenue data not matched 0.197 0.398 0.109 0.312
No debt outstanding 0.398 0.489 0.048 0.214
Debt – 1st quartile 0.400 0.490 0.785 0.411
Debt – 2nd quartile 0.005 0.068 0.050 0.217
Debt – 3rd quartile 0.000 0.022 0.007 0.081
Debt – 4th quartile 0.000 0.009 0.001 0.035
Debt data not matched 0.197 0.398 0.109 0.312
Budget growth ⩽−7% 0.193 0.395 0.156 0.363
Budget growth −7%–+18% 0.190 0.393 0.292 0.455
Budget growth ⩾ +18% 0.182 0.386 0.267 0.442
Budget data not matched 0.434 0.496 0.286 0.452
State issuers per 100,000 residents State market share of par value
Any (Var/Put/Corp) 0.263 0.194 0.229 0.119
Variable rate 0.176 0.149 0.149 0.106
Put option 0.142 0.121 0.118 0.093
Corporate 0.161 0.152 0.110 0.082
Derivative 0.065 0.080 0.025 0.039
n 2,061,024 130,112

Note: Data are from the Mergent Municipal Bond Securities Database and the Census of Governments

Issuer characteristics and use of bonds with innovations

Any (Var/Put/Corp) Variable rates Put option Corporate Derivative
State 0.562*** 0.506*** 0.477*** 0.542*** 0.230***
(0.038) (0.037) (0.039) (0.049) (0.031)
County 0.030*** 0.023*** 0.018*** 0.020*** 0.000
(0.002) (0.002) (0.001) (0.001) (0.001)
Special district 0.001*** 0.001* 0.000* 0.001*** 0.001***
(0.000) (0.000) (0.000) (0.000) (0.000)
School district −0.008*** −0.006*** −0.004*** −0.004*** 0.001**
(0.000) (0.000) (0.000) (0.000) (0.000)
Revenue – 2nd quartile −0.001*** −0.001*** −0.001*** −0.000 −0.000*
(0.000) (0.000) (0.000) (0.000) (0.000)
Revenue – 3rd quartile 0.001*** 0.000* 0.000 0.001*** −0.000
(0.000) (0.000) (0.000) (0.000) (0.000)
Revenue – 4th quartile 0.013*** 0.009*** 0.006*** 0.006*** 0.002***
(0.001) (0.000) (0.000) (0.000) (0.000)
Revenue, debt not matched 0.005*** 0.003*** 0.002*** 0.003*** 0.001***
(0.000) (0.000) (0.000) (0.000) (0.000)
Debt – 1st quartile 0.002*** 0.001*** 0.001*** 0.001*** 0.000***
(0.000) (0.000) (0.000) (0.000) (0.000)
Debt – 2nd quartile 0.211*** 0.169*** 0.145*** 0.108*** 0.112***
(0.010) (0.009) (0.008) (0.008) (0.006)
Debt – 3rd quartile 0.335*** 0.302*** 0.288*** 0.216*** 0.373***
(0.044) (0.041) (0.040) (0.042) (0.039)
Debt – 4th quartile 0.480*** 0.513*** 0.529*** 0.430*** 0.669***
(0.067) (0.056) (0.045) (0.049) (0.047)
Budget growth ⩽−7% −0.001** −0.001** −0.001*** −0.000* 0.000
(0.000) (0.000) (0.000) (0.000) (0.000)
Budget growth ⩾+18% 0.001** 0.001*** 0.000* −0.000 0.001***
(0.000) (0.000) (0.000) (0.000) (0.000)
Budget data not matched 0.000 0.000 0.000 −0.000 0.000
(0.000) (0.000) (0.000) (0.000) (0.000)
State issuer/100K pop – 1 yr lag 0.005*** 0.007*** 0.006*** 0.004*** 0.012***
(0.000) (0.001) (0.001) (0.000) (0.001)
State issuer/100K pop – 2 yr lag 0.002*** 0.003*** 0.004*** 0.001*** 0.004***
(0.000) (0.001) (0.001) (0.000) (0.001)
State issuer/100K pop – 3 yr lag 0.003*** 0.002*** 0.001** 0.002*** 0.002**
(0.000) (0.000) (0.000) (0.000) (0.001)
State fixed effects Y Y Y Y Y
Year fixed effects Y Y Y Y Y
Constant 0.000 0.002 0.002 0.001 −0.002***
(0.001) (0.001) (0.001) (0.001) (0.000)
R2 0.119 0.122 0.123 0.108 0.140
n 2,061,024 2,061,024 2,061,024 2,061,024 2,061,024

Notes: Data are from the Census of Governments and the Mergent Municipal Bond Securities Database. The dependent variable is an indicator of any issuance utilizing the innovation. Observations are issuer-years. Standard errors are clustered by the issuer. *p<0.05; **p<0.01; ***p<0.001

Issuer characteristics and percent of par value with innovations

Any (Var/Put/Corp) Variable rates Put option Corporate Derivative
State 12.726*** 9.581*** 5.785*** 6.809*** −1.108*
(2.233) (1.808) (1.361) (1.701) (0.474)
County 6.249*** 3.984*** 2.974*** 4.106*** −0.027
(0.480) (0.375) (0.313) (0.373) (0.053)
Special district 7.257*** 4.018*** 2.802*** 4.956*** 0.734***
(0.596) (0.426) (0.345) (0.457) (0.090)
School district −2.971*** −2.913*** −1.869*** −2.068*** 0.107**
(0.217) (0.158) (0.130) (0.168) (0.035)
Revenue – 2nd quartile 1.739 0.808 0.556 1.460 0.369*
(1.631) (1.202) (1.000) (1.415) (0.157)
Revenue – 3rd quartile 1.737 0.602 0.245 0.773 0.494***
(1.527) (1.133) (0.944) (1.319) (0.132)
Revenue – 4th quartile 3.358* 1.565 0.892 1.643 0.754***
(1.551) (1.153) (0.957) (1.329) (0.138)
Revenue, debt not matched 4.556** 1.563 1.197 3.312* 0.835***
(1.539) (1.149) (0.961) (1.354) (0.146)
Debt – 1st quartile −2.657*** −1.154** −0.825* −2.162*** −0.170*
(0.559) (0.423) (0.342) (0.430) (0.071)
Debt – 2nd quartile 3.520*** 3.953*** 3.386*** −0.540 1.606***
(0.889) (0.673) (0.561) (0.685) (0.162)
Debt – 3rd quartile 2.055 2.604 2.402 −0.816 3.633***
(2.131) (1.599) (1.328) (1.533) (0.668)
Debt – 4th quartile 4.696 2.224 2.840 4.037 3.664***
(3.724) (2.701) (1.858) (3.143) (0.556)
Budget growth ⩽−7% 0.057 0.189 −0.024 −0.126 0.106*
(0.235) (0.188) (0.157) (0.183) (0.047)
Budget growth ⩾ +18% −0.230 0.006 −0.043 −0.290 0.126***
(0.201) (0.164) (0.141) (0.151) (0.035)
Budget data not matched −0.370 0.655* 0.549* −0.489 0.090
(0.369) (0.298) (0.248) (0.287) (0.055)
Market share in state – 1 yr lag 7.399*** 6.745*** 6.798*** 6.621*** 3.137***
(0.821) (0.709) (0.732) (1.130) (0.646)
Market share in state – 2 yr lag 6.141*** 2.998*** 3.494*** 6.205*** −0.476
(0.808) (0.690) (0.711) (1.042) (0.494)
Market share in state – 3 yr lag 2.121** 1.821** 0.975 4.431*** 1.232**
(0.783) (0.664) (0.722) (1.071) (0.437)
Yields, no innovation 8.669*** 3.725*** 3.900*** −3.237*** 0.227***
(0.274) (0.191) (0.184) (0.245) (0.037)
Spread, innovation 4.688*** 0.026 1.065*** −3.586*** 0.137***
(0.141) (0.063) (0.101) (0.182) (0.015)
State fixed effects Y Y Y Y Y
Year fixed effects Y Y Y Y Y
Constant −37.494*** −15.325*** −16.611*** 20.737*** −2.128***
(2.438) (1.788) (1.598) (2.090) (0.279)
R2 0.144 0.096 0.080 0.085 0.032
n 130,111 130,111 130,111 122,199 130,111

Notes: Data are from the Census of Governments and the Mergent Municipal Bond Securities Database. The dependent variable is the percent of par value utilizing the innovation. Observations are issuer-years. Standard errors are clustered by the issuer. *p<0.05; **p<0.01; ***p<0.001

Time trends for adoption of bonds with innovations

Any (Var/Put/Corp) Variable rates Put option Corporate Derivative
By type of jurisdiction
(Cities) Year −0.20** −0.30*** −0.11* 0.33*** 0.08***
(0.06) (0.05) (0.05) (0.04) (0.01)
Schools×Year 0.23*** 0.25*** 0.11* −0.24*** 0.08**
(0.07) (0.06) (0.05) (0.05) (0.02)
Spec. Dist.×Year 0.71*** 0.43** 0.45** 0.90*** 0.46***
(0.19) (0.16) (0.15) (0.15) (0.07)
County×Year 0.05 −0.12 0.30* 0.84*** 0.01
(0.17) (0.16) (0.14) (0.14) (0.04)
State×Year 2.26*** 1.70*** 1.64*** 1.53*** 0.90***
(0.35) (0.33) (0.27) (0.30) (0.12)
R2 0.08 0.07 0.05 0.04 0.02
n 101,714 101,714 101,714 101,714 101,714
By quartile of debt outstanding
(1st quartile) Year −0.34*** −0.39*** −0.17*** 0.20*** 0.10***
(0.05) (0.04) (0.03) (0.03) (0.01)
2nd quartile×Year 0.72** 0.62* 0.73** 0.50** 0.68***
(0.27) (0.25) (0.22) (0.19) (0.11)
3rd quartile×Year 1.97*** 1.50*** 1.39*** 1.51*** 1.33***
(0.53) (0.45) (0.40) (0.39) (0.32)
4th quartile×Year 1.93* 1.02 1.20* 2.05*** 1.39***
(0.98) (0.78) (0.57) (0.58) (0.20)
R2 0.04 0.05 0.04 0.01 0.04
n 87,696 87,696 87,696 87,696 87,696
By quartile of debt issued in the last 3 years
(1st quartile) Year 0.26** −0.08 0.02 0.61*** 0.07***
(0.10) (0.08) (0.07) (0.07) (0.01)
2nd quartile×Year −0.51*** −0.24** −0.12 −0.26** 0.04*
(0.11) (0.09) (0.08) (0.08) (0.02)
3rd quartile×Year 0.44 0.65** 1.06*** 0.51* 0.99***
(0.25) (0.23) (0.20) (0.20) (0.12)
4th quartile×Year 1.80*** 1.18** 1.16** 1.41*** 1.99***
(0.47) (0.43) (0.40) (0.38) (0.35)
R2 0.04 0.05 0.04 0.02 0.05
n 105,260 105,260 105,260 105,260 105,260

Notes: The dependent variable is the share of par value. Observations are issuer-years. “Quartile” indicates the quartile of total debt outstanding or the quartile of the total debt issued in the preceding three years. The first quartile (omitted) indicates the least debt and fourth quartile indicates the most debt. Data are from the Census of Governments and the Mergent Municipal Bond Securities Database. Standard errors are clustered by the issuer. *p<0.05; **p<0.01; ***p<0.001

Notes

1.

References to municipal bonds as “boring” are very common in the popular and financial press. For examples, see Sullivan (2011), Deener (2016) and Foster (2017). However, these informal statements do not make it clear whether they are referring to lack of innovation, appreciation, volatility or something else.

2.

There have been many studies of the adoption of technology or nonfinancial policies by state and municipal governments. For examples, see Berry and Berry (2014) and Moon et al. (2014).

3.

There are 262,069 observations dated before 1992, but the annual counts are less than half the post-1992 counts. The pre-1992 observations may only represent a subset of all bonds issued, and the selection into that sample is unknown. Therefore, the analysis is limited to years 1992 and later.

4.

Other published annual totals include only long-term bonds, so I excluded short-term bonds, notes and other types of securities from the Mergent data before creating annual totals for comparison.

5.

These figures are available publicly through the Securities Industry and Financial Markets Association (SIFMA) and the Bond Buyer. SIFMA Statistics www.sifma.org/resources/research/us-municipal-issuance/ (accessed January 25, 2018). Bond Buyer Annual Municipal Debt Sales www.bondbuyer.com/marketstatistics/search_amdt.html (accessed January 25, 2018).

6.

State and Local Government Finance Historical Data www.census.gov//govs/local/historical_data.html (accessed January 25, 2018).

7.

In total, 5 percent of the issuers could not be matched with the COG data, so they are excluded from parts of the analysis that require COG measures. They are included for all calculations that do not require merged observations.

8.

Identifying an innovation in the first year of the data might be possible if one has reliable external information. Tufano declared one security in the first year of his data to be an innovation based on a literature search and interviews with investment bankers. The first appearance of the other 57 innovations was observable in the third or later years of his data.

9.

Even if we grouped together uncommon features that seem similar, there are only a couple potential popularized clusters. The combination of all housing-related bonds increased from a 3.9 percent market share to a 10.1 percent market share. A combination of two types of remarketed and convertible bonds rises from a 3.0 to 10.6 percent market share.

10.

These counts exclude the features based on the “use of proceeds” variable. Every observation has to have some intended use recorded in the data, and the fine gradations (58 categories) keep almost all of them below a 5 percent share. If these are included, nearly all securities appear to have at least one uncommon feature.

11.

This is the number of days which the holder of a bond must be notified in advance of a redemption by the obligated party. The standard is 30 days.

12.

Raising the popularization threshold to 15 percent would not highlight any additional innovations or exclude any of the four focal features. If we lowered the threshold to 5 percent, variable rate securities would not be an innovation because they already had a 5 percent market share in 1993.

13.

Mergent provided the following list of debt types that it instructs its analysts to categorize as derivatives: residuals, floaters, floats, residual interest tax exempt securities (RITES), putters, drivers, residual option certificates (ROCs), residual option longs (ROLs), eagles, spears, lifers, P-floats, residual interest bonds (RIBS), select auction variable rate securities (SAVRS), tender option bonds and TR receipts.

14.

As described in Stewart and Smith (2012), the regulatory change that impacted auction rate securities was a statement by the Securities Exchange Commission (SEC) that ARS should be disclosed as long-term securities. Dealers had been representing them as short-term cash equivalent securities. When corporate treasurers had to recognize the securities as long-term instruments, demand for them dropped.

15.

The par value of derivatives is not included in Figure 6. Equivalent figures for derivatives are available upon request.

16.

As an alternative to the total debt outstanding measure, we could also group issuers by how much debt they have issued in the last three years. Time series for these groupings would lead to the same conclusion. The most active market participants were the ones who shifted more of their market share to securities with innovative features.

17.

In the models in Tables VII and VIII, the observations are issuer-years. The municipal governments that do not issue could be merged with a uniform average of yields over the year. However, this has no within-year variation, so it is absorbed by the year fixed effect. In Table VII, yields are observed in the three months preceding an issuance and then aggregated to an annual average. However, in the presence of year fixed effects, the coefficient on the yield and spread measures are identified only by the within-year variation created by issuers coming to market in different months. This does not enable us to contrast high- and low-interest rate environments.

References

Alderson, M.J. and Fraser, D.R. (1993), “Financial innovations and excesses revisited: the case of auction rate preferred stock”, Financial Management, Vol. 22 No. 2, pp. 61-75.

Beck, T., Chen, T., Lin, C. and Song, F.M. (2016), “Financial innovation: the bright and the dark sides”, Journal of Banking and Finance, Vol. 72, pp. 28-51.

Bergstresser, D. and Orr, P. (2014), “Direct bank investment in municipal debt”, Municipal Finance Journal, Vol. 35 No. 1, pp. 1-23.

Berry, F.S. and Berry, W.D. (2014), “Innovation and diffusion models in policy research”, in Sabatier, P.A. and Weible, C.M. (Eds), Theories of the Policy Process, 3rd ed., Westview Press, Boulder, Colorado, pp. 307-362.

Cavallaro, L. (2012), Hospital Bond Analysis, John Wiley & Sons, Hoboken, NJ, pp. 845-859.

Cestau, D., Green, R.C. and Schurhoff, N. (2013), “Tax-subsidized underpricing: the market for Build America Bonds”, Journal of Monetary Economics, Vol. 60 No. 5, pp. 593-608.

Deener, W. (2016), “Three reasons why boring ol’ municipal bonds might be the way to go”, Dallas News, July 22, available at: www.dallasnews.com/business/personal-finance/2016/07/22/investments-might-boring-stayed-strong-markets-tanked

Denison, D.V. and Gibson, J.B. (2013), “A tale of market risk, false hope, and corruption: the impact of adjustable rate debt on the Jefferson County, Alabama Sewer Authority”, Journal of Public Budgeting, Accounting & Financial Management, Vol. 25 No. 2, pp. 311-345.

Felstein, S., Fabozzi, F.J. and Fabozzi, T. (1995), “Municipal bonds”, in Fabozzi, F.J. and Fabozzi, T.D. (Eds), The Handbook of Fixed Income Securities, 4th ed., Irwin Professional Publishing, New York, NY, pp. 155-185.

Financial Crisis Inquiry Commission (2011), “The financial crisis inquiry report: final report of the national commission on the causes of the financial and economic crisis in the United States”, Public Affairs, Financial Crisis Inquiry Commission, Washington, DC.

Finnerty, J.D. (1988), “Financial engineering in corporate finance: an overview”, Financial Management, Vol. 17 No. 4, pp. 14-33.

Finnerty, J.D. (1992), “An overview of corporate securities innovation”, Journal of Applied Corporate Finance, Vol. 4 No. 4, pp. 23-39.

Finnerty, J.D. and Emery, D.R. (2002), “Corporate securities innovation: an update (digest summary)”, Journal of Applied Finance, Vol. 12 No. 1, pp. 21-47.

Foster, M. (2017), “Municipal bonds: Wall Street’s hidden 6.5% income secret”, Forbes, September 6, available at: www.forbes.com/sites/michaelfoster/2017/09/06/municipal-bonds-wall-streets-hidden-6-5-income-secret/#3304280017eb

Frame, W.S. and White, L.J. (2004), “Empirical studies of financial innovation: lots of talk, little action?”, Journal of Economic Literature, Vol. 42 No. 1, pp. 116-144.

Gennaioli, N., Shleifer, A. and Vishny, R. (2012), “Neglected risks, financial innovation, and financial fragility”, Journal of Financial Economics, Vol. 104 No. 3, pp. 452-468.

Goyal, V.K., Gollapudi, N. and Ogden, J.P. (1998), “A corporate bond innovation of the 90s: the clawback provision in high-yield debt”, Journal of Corporate Finance, Vol. 4 No. 4, pp. 301-320.

Hildreth, W.B. and Zorn, C.K. (2005), “The evolution of the state and local government municipal debt market over the past quarter century”, Public Budgeting and Finance, Vol. 25 No. 4s, pp. 127-153.

Kidwell, D.S. and Rogowski, R.J. (1983), “Bond banks: a state assistance program that helps reduce new issue borrowing costs”, Public Administration Review, Vol. 43 No. 2, pp. 108-113.

Lebenthal, J., Kanner, B. and Volcker, P. (2006), Confessions of a Municipal Bond Salesman, Wiley, Hoboken, NJ.

Liu, G. and Denison, D.V. (2014), “Indirect and direct subsidies for the cost of government capital: comparing tax-exempt bonds and Build America Bonds”, National Tax Journal, Vol. 67 No. 3, pp. 569-593.

Luby, M.J. (2012), “Federal intervention in the municipal bond market: the effectiveness of the Build America Bond program and its implications on federal and subnational budgeting”, Public Budgeting and Finance, Vol. 32 No. 4, pp. 46-70.

Luby, M.J. and Kravchuk, R.S. (2013), “An historical analysis of the use of debt-related derivatives by state governments in the context of the great recession”, Journal of Public Budgeting, Accounting & Financial Management, Vol. 25 No. 2, pp. 276-310.

Miller, M.H. (1986), “Financial innovation: the last twenty years and the next”, Journal of Financial and Quantitative Analysis, Vol. 21 No. 4, pp. 459-471.

Mincke, B.D. (2012), How to Analyze Higher Education Bonds, John Wiley & Sons, Hoboken, NJ, pp. 1055-1076.

Moon, M.J., Lee, J. and Roh, C.-Y. (2014), “The evolution of internal IT applications and e-Government studies in public administration: research themes and methods”, Administration & Society, Vol. 46 No. 1, pp. 3-36.

Mysak, J. (2012), Encyclopedia of Municipal Bonds: A Reference Guide to Market Events, Structures, Dynamics, and Investment Knowledge, Bloomberg Financial, Wiley, Hoboken, NJ.

Singla, A. and Luby, M.J. (2014), “A descriptive analysis of state government debt-related derivatives policies”, Public Budgeting & Finance, Vol. 34 No. 2, pp. 105-125.

Stewart, L.J. and Cox, C.A. (2008), “Debt-related derivative usage by US state and municipal governments and evolving finance reporting standards”, Journal of Public Budgeting, Accounting & Financial Management, Vol. 20 No. 4, pp. 439-456.

Stewart, L.J. and Smith, P.C. (2012), “The 2008 auction rate securities market collapse and US nonprofit health systems”, Qualitative Research in Financial Markets, Vol. 4 No. 1, pp. 68-83.

Sullivan, P. (2011), “For investing big bonuses, boring is in”, New York Times 160(55293), January 21, p. 6.

Temel, J.W. (2011), The Fundamentals of Municipal Bonds, Wiley Finance, Wiley, Hoboken, NJ.

Tufano, P. (1989), “Financial innovation and first-mover advantages”, Journal of Financial Economics, Vol. 25 No. 2, pp. 213-240.

Tufano, P. (2003), “Financial innovation”, in Constantinides, G.M., Harris, M. and Stulz, R.M. (Eds), Handbook of the Economics of Finance, Vol. 1, Elsevier, Amsterdam, pp. 307-335.

Wooldridge, P. (2016), “Comparison of BIS derivatives statistics”, 8th IFC Conference, Hong Kong, September 8-9.

Corresponding author

Stephan David Whitaker can be contacted at: stephan.whitaker@clev.frb.org