Economic fundamentals, policy responses, and state-level municipal bond sensitivity to COVID-19 prevalence

Abstract

This paper conducts a state-by-state analysis of the financial impact of the COVID-19 pandemic on the U.S. municipal bond market. Using panel regressions and state-by-state regressions, we find that the prevalence rates of the COVID-19 virus negatively impacted the aggregate performance of municipal bonds. The study also explored whether the disparities in the economic fundamentals of U.S. states, as well as the COVID-19 mitigation policies employed by each state, can explain the sensitivity of the state’s municipal bond to its COVID-19 prevalence rates. States with more desirable economic fundamentals and robust COVID-19 mitigation policies appeared to have higher COVID-19 sensitivities than states that do not. This finding may be due to a baseline effect, in which the higher levels of economic activities in these states also make them more susceptible to the deleterious effects of the stronger mitigation policies enacted by them.

1. Introduction

The devastating human and economic costs of the COVID-19 pandemic have prompted many researchers to explore how the healthcare crisis has affected capital markets and economies around the world. Indeed, in the ensuing periods, there has been a spurt of literature that has examined how the crisis has affected the performances of the stock markets, the commodity markets, the foreign exchange (FX) markets, the cryptocurrency markets, and the bond markets. Contessi and De Pace (2020) find evidence of general stock market collapse and the transmission of shocks across 18 global capital markets in the periods following the onset of the COVID-19 pandemic. Ashraf (2020) examines the stock markets’ reaction to three types of government actions in 77 different countries. The government actions are grouped into three major categories: social distancing measures which include policies dealing with the closure of schools, workplaces, parks, public transportation, etc., containment measures which include public awareness campaigns, testing, and quarantine policy, and income support policies which include government financial assistance to households and businesses. His findings suggest that the government containment measures, and income support packages mostly had positive effects on the stock markets, while the social distancing measures had both positive and negative economic effects. Corbet et. al (2020b) find support for “flight to safety” in the same period, with a significantly high correlation between the Chinese stock market returns and the value of Bitcoin and Gold. Ashraf (2021) presents evidence that the intensity of the negative effect of COVID-19 on stock markets around the world can be partly explained by the national level of uncertainty avoidance. Uncertainty avoidance is defined as the degree to which members of a culture feel uncomfortable with uncertainty/unstructured situations. Ding et al. (2021) established that firms with stronger pre-COVID-19 financial conditions, lower international exposure, stronger corporate social responsibility (CSR) ranking, less entrenched management, and less institutional ownership experienced smaller stock price decline during the pandemic. Corbet et al. (2020a) present evidence of volatility spillover from oil to renewables and coal in the period surrounding the unprecedented crash of the West Texas Intermediate (WTI) oil price to a negative price in April 2020.

A number of authors also focused on assessing the reaction of the bond markets to COVID-19 related policy decisions. Nozawa and Qiu (2021) find disparities in the reactions of corporate bond spreads to the Federal Reserve monetary policy announcements. In particular, they found that credit spreads on investment-grade bonds decreased significantly after the March 23rd, 2020 Federal Reserve announcement (of a range of new programs to promote the stability of the financial system) while spreads on high-yield bonds did not.1 In contrast, spreads on both investment grade and high yield bonds fell after the April 9th announcement of the Municipal Liquidity Facility (MLF) aimed at purchasing $500 billion in high-quality notes directly from states and local governments. Bordo and Duca (2021) provide an overview of the motivation for various Federal Reserve policy tools that were employed to combat the adverse effects of the COVID-19 pandemic. The authors provide a detailed explanation of the new credit-easing policies that were primarily intended to dampen the feedback mechanisms that would otherwise magnify the downturn and hinder a subsequent recovery. Additionally, they provide an assessment of the impact of the new policy tools and address the risks they might pose. For the municipal bond financing, they discussed the impact of the Fed policy regarding the Municipal Liquidity Facility, as well as analyzed the studies that examined the impact of other fiscal policies such as the Paycheck Protection Program (PPP) on state and local levels of employment and conclude that the combination of federal fiscal support to states, along with the creations of the PPP and the MLF programs may have prevented municipal debt distress from escalating and posing a systemic risk to the financial system. He et al. (2021) find a substantial increase in Treasury yield over similar-maturity overnight index swap rate for secured financing at the onset of the COVID-19 pandemic. This finding implies the presence of an inconvenience yield, which may be due to the concern amongst investors that the safe-haven status of U.S. Treasuries could be in jeopardy. Indeed, they found that large owners of Treasuries substantially reduced their holdings during March 2020. This created a supply shock of U.S. Treasuries which the intermediary sector struggled to absorb. Li and Lu (2020), using an event study methodology and county-level municipal bond data from January 1, 2020, to April 30, 2020, examine the effect of COVID-19 on municipal bond performance and new issues. In general, they find that the average offering yield increased while the number of new issues drops when county-level COVID-19 case and death counts increase, with bond investors paying more attention to COVID-19 virus-related mortality than to other factors. Finally, Goodell (2020) provides a comprehensive literature survey on the impact of COVID-19 on the financial sector and suggests that the COVID-19 pandemic may have had an important impact on the functioning of the financial sector.

Overall, the growing consensus in the literature is that the COVID-19 pandemic had a negative effect on capital markets around the world. While the focus of many of the recently published works has been on the impact of the pandemic on businesses and capital markets, it is important to note that the bulk of the actions taken to mitigate the spread of the virus and a significant share of its human and economic costs was borne by governments at all levels. For example, aggregate total taxes collected across the 50 U.S. states and the District of Columbia fell by 2.5% between 2019 and 2020 ($1.093 trillion to $1.066 trillion). A substantial component of this decline was due to a steeper decline in income tax collections, which fell by 6.5% between 2019 and 2020 ($471.6 to $440.9 billion). In reality, there were significant variations in the impact of the pandemic on tax receipts across U.S. states. Thirty-six states (including the District of Columbia) saw a decline in total tax receipts, with an average decline of approximately 4.9% and the highest decline of 25.9% for the state of Alaska. Thirty-six out of the 47 states that collect income taxes, experienced a decline in income tax receipts for 2020, with an average decline of approximately 10.68% and the highest decline of 51.09% for the state of Alaska. Furthermore, a recent survey of 900 municipalities by the National League of Cities (NLC) shows that 90% of municipalities experienced a decrease in revenue while 76% took on additional expenses. The same survey revealed that cities on average experienced a 21% decline in revenue in 2020, while expenditures increased by 17% during the same period.2 The deteriorating fiscal condition of state and local municipalities during the pandemic underscores a potential source of credit risk to municipal bonds as many municipalities rely primarily on taxes and revenue from projects to fund these obligations.

Given all the compelling facts stated above, it is surprising that no study has examined in detail the effect of the pandemic on the finances of U.S. states and local governments. The goal of this research is to fill this void. In this paper, we conduct a comprehensive analysis of the impact of the COVID-19 pandemic on the finances of U.S. states and local governments. In particular, the study conducts a state-by-state analysis of the linkages between the prevalence rates of the COVID-19 virus in each state and the financial performances of municipal bonds issued in the state. Municipal bonds are debt obligations issued by states, local governments, and other tax-exempt entities to fund public projects such as schools, roads, hospitals, and airports. They allow states and local governments to finance budgetary outlays using expected future cash flows. Consequently, their performances are intricately tied to the future economic well-being of the public entities that issued these bonds. Indeed, Grigoris (2019) found that the municipal yield curves contain information about the future local macroeconomic and financial market outcomes.3 Thus, an examination of the impact of the COVID-19 pandemic on U.S. municipal bond returns and yields could yield further insights into the localized costs of the pandemic across the 50 U.S. states and the District of Columbia.

Besides the fact that they provide efficient tax shelters for certain classes of investors, municipal bonds constitute a major sector of the fixed income market. With approximately $4.0 trillion of municipal bonds outstanding, it is the fourth largest sector of the domestic fixed income market (behind Treasury, mortgage-related, and corporate bond markets in descending order). Unlike the equity markets which operate primarily through centralized exchanges, municipal bonds trade mostly over the counter through a network of dealers who maintain inventories of municipal bonds. Before the pandemic, the financial health of many states and local governments was in relatively good condition and most municipal bonds were considered safe investments. Investor interests in municipal bonds were high with municipal bond funds posting record inflows in 2019 (McDevitt & Watson, 2020). Indeed, in that same year, approximately $2.9 trillion of municipal bonds were traded. About 51% of this notional amount was for the purchase of municipal bonds by investors, while about 31% and 18% of the same amount were for investors’ sales and inter-dealer trades, respectively. Furthermore, the primary market for municipal bonds reached a ten-year high in 2020, when 14 thousand new municipal bonds with a notional value of $516 billion were issued, as states and municipalities scrambled to plug the precipitous decline in their cash inflows. The ensuing liquidity crisis triggered a jump in municipal bond yield, which increased by more than 200 basis points in March 2020. Besides the above, the effective spread on municipal bond yield which had also been on the decline in the four years prior to 2020, suddenly jumped from 55 basis points to 97 basis points.4 In response to the ensuing crisis, the Federal Reserve expanded the scope of its Money Market Liquidity Facility to allow for the purchase of municipal bonds in March 2020. The Federal Reserve also established the Municipal Liquidity Facility in April 2020, to serve as a direct source of emergency financing to states and municipalities during the crises. The rapid ascent of municipal bond yields which was observed in March 2020 was a short-lived phenomenon, as yields quickly collapsed to levels that were lower than the values seen before the advent of the pandemic. Perhaps due to the combined effects of the Federal Reserve programs and the enactment of the Coronavirus State and Local Fiscal Recovery funds (through the CARES ACT).

In the realm of academic research, interest in municipal bonds is low because unlike the equity and corporate bond markets, the municipal bond market is less transparent with respect to issuer-specific information due to the lower disclosure requirements for municipal bonds issuers. The municipal bond market is also relatively more illiquid than the equity and bond markets, due to the relatively low volume of trades that occur in municipal securities once they are issued. Harris & Piwowar (2006) find that municipal bond trades are substantially more expensive than the cost of similar-sized trades in equity due to the limited availability of bond market data for muni bonds. Schwert (2017) finds that yield spreads on municipal securities primarily account for default risk, although defaults in municipal bonds are inherently rare events. 5 The upshot of these factors is that the magnitude of the information asymmetry which exists in the municipal bond market may be more pronounced than those found in other financial markets.6

The first confirmed case of the COVID-19 virus was identified in Washington State on January 20th, 2020. Since then, the virus has spread to all 50 U.S. states, the District of Columbia, and U.S. territories, with over 30 million confirmed cases and over 600 thousand deaths in the U.S. alone, at the time of the writing of this manuscript. While the aggregate toll of the virus on the U.S. is indeed staggering, it is important to note that there was a wide variation in the human and economic tolls of the virus across the 50 states and the District of Columbia.7 Large disparities were also apparent in the policy responses enacted by states and local governments to mitigate the impact of the virus. Furthermore, economic fundamentals vary considerably across U.S. states8 and so does the depth of the municipal bond market activity by each state.9 There is also a wide variation in the types of municipal bonds issued by public entities in each state.10 In all, the above provide compelling motivations for researchers to examine whether the pandemic had impacted the market performances of the municipal bonds issued in each state and whether the policy actions and economic fundamentals of each state can explain the sensitivity of the state's municipal bond performance to COVID-19 prevalence.

In this study, an aggregate and state-by-state scrutiny of the municipal bond market are conducted to examine whether the daily performance of each state’s municipal bond indices can be explained by the prevalence of COVID-19 in that state. Furthermore, the study controls for the possibility that performances of municipal bond indices could be explained by other factors by including four other covariates in the econometric specifications utilized in this study. The econometric specifications employed to test for COVID-19 effects in the municipal bond index and bond yield performances include panel regressions, pooled regressions, cross-sectional regressions, and linear discriminant analysis.

In the results section, the paper presents evidence that alludes to the fact that COVID-19 prevalence levels have a significant negative impact on the performance of municipal bonds in general. The negative effects uncovered were prominent in more than half of the U.S. states (including the District of Columbia). The result is robust, after controlling for other determinants of state-level municipal bond fund performances including the aggregate stock market performance and the aggregate performance of investment quality and high-yield municipal bonds. The results from the study shed pertinent information for both public and private stakeholders. From the public policy point of view, the findings from this research will clarify how pandemics shocks are propagated within and across markets. For example, policymakers should be aware that the economic cost of pandemic shocks is not only limited to the equity and corporate bond markets. Indeed, the results obtained suggest that while the performances of municipal bonds are for the most part uncorrelated to the equity markets, they are by no means immune to the pandemic. From the investor point of view, the findings from this research will provide additional insights into how investors react to changes in the investment opportunity set as pandemic-related shocks flow into the municipal bond market. In particular, the results from this research will be enormously useful for portfolio allocation, risk management, asset pricing, and cross-market hedging. For example, if municipal bond performances display relatively lower exposure to pandemics than other markets, then they may be attractive options for investors looking for yields during flight-to-safety episodes.

The rest of this paper is organized as follows. In Section II, state-level municipal bond and COVID-19 data, as well as other financial state variables used in the econometric model are described. Section III presents the panel regression models used to determine the impact of the COVID-19 pandemic on the returns and yields of municipal bonds. The results from estimating the various econometric specifications are subsequently discussed. Section IV presents the linear regression specification used to conduct the state-by-state assessment of the COVID-19 effect on municipal bond performances. Section V presents the roles of economic fundamentals and policy responses in explaining the COVID-19 sensitivities of municipal bond performances. Section VI contains concluding remarks.

2. Data description

The data utilized in this study is comprised of primary and secondary data obtained from March 2020 to January 2021. The primary data includes the daily percent changes in the Bloomberg Barclay’s municipal bond total return indices and the Standard and Poor’s (S&P) municipal bond yield indices for all 50 U.S. states and the District of Columbia. The state-level data described above, as well as daily values of the financial covariate variables described in the subsequent paragraph of this section, were obtained from Bloomberg Professional Service.

Daily COVID-19 data for all 50 U.S. states and the District of Columbia were obtained from the COVID Tracking Project.11 The data includes the number of COVID-19 tests, the number of positive COVID test results, the number of COVID deaths, hospitalizations, and patients in intensive care units (ICU) to mention a few. Publicly available COVID-19 data suffers from asynchronous biases due to the delay in the reporting of COVID statistics by responsible parties and the periodic revision of existing data. To reduce the impact of these factors on the analyses, the 30-day moving average of the daily percent change in the number of positive cases is calculated. 12 The 30-day moving average of the daily percent change in the number of positive cases is a robust measure of the prevalence of COVID cases for the following reasons. First, COVID testing data is a lagging indicator of the current levels of virus prevalence in an area because cases are confirmed after exposure has occurred. Furthermore, the turnaround time for the PCR test, which is the most widely accepted method of detecting COVID cases, could be as high as 72 hours.13 Hence, in this study, we focus on the relationship between the 30-day moving average of the daily changes in the number of positive cases, and the state’s municipal bond performances.

It is well established in the literature that financial markets tend to comove more significantly during periods of uncertainty.14 Thus, we control for the likelihood that state-level municipal bond performances could be explained by events in other financial markets by including the daily logarithmic returns on the S&P 500, the Bloomberg Barclays AAA Municipal Bond Index, the Bloomberg Barclays High Yield Municipal Bond Index, and the daily percent changes in the value of the CBOE Volatility (VIX) index.

Table 1 reports the summary statistics obtained from the state-level data. In Panel A, summary statistics of the municipal bond returns and yields from all 50 states and the District of Columbia are reported. The same statistics for the COVID-19 variable and the other four financial covariates are also reported in Panel B. The average daily return on the municipal bond indices is close to zero. Changes in the average yields for the municipal bond yield index are also close to zero. In comparison to the S&P 500 which had an average daily return of 0.095%. The risks of municipal bonds are also generally lower than the risk of the broader stock market as shown by the values of the standard deviations for the bond indices which were found to be much lower than the standard deviation of the S&P 500. Turning our attention to other properties of the distribution. The returns of the municipal bond indices for approximately half of the states are negatively skewed. The same statistics for the municipal bond yield are positively skewed for the entire 50 states and the District of Columbia. Both the municipal bond returns and municipal bond yields display significant excess kurtosis. Further evidence in support of these statistics can be found in the p-values of the Jarque-Bera tests of normality, which rejects the null that all municipal bond return and yield series are normally distributed. The significance of the results of the Durbin-Watson test implies the presence of serial correlations in the error distribution. Taken together, the statistical properties of the municipal bond returns and yields suggest that we test our null hypotheses using heteroscedasticity robust standard errors.

Table 1.

Summary Statistics.

Panel A: Municipal Bond Performances
U.S. States Municipal Bond Return Index											U.S. State Municipal Bond Yield Index
State	Mean	Std. Dev	Min	Max	SK	KU	J-B (MSL)	DW (MSL)	ADF (MSL)		Mean	Std. Dev	Min	Max	SK	KU	J-B (MSL)	DW (MSL)	ADF (MSL)
AK	0.011	0.41	−2.57	2.96	−0.17	27.29	0.00	0.00	0.00		−0.12	4.72	−25.79	31.58	1.71	20.96	0.00	0.00	0.00
AL	0.014	0.46	−2.57	3.00	−0.69	25.90	0.00	0.00	0.00		−0.13	3.88	−21.02	28.30	1.90	22.91	0.00	0.00	0.00
AR	0.013	0.37	−2.30	2.57	−0.45	27.07	0.00	0.00	0.00		−0.06	3.74	−24.38	23.71	0.42	23.89	0.00	0.00	0.00
AZ	0.010	0.46	−2.56	3.17	−0.09	25.09	0.00	0.00	0.00		−0.07	3.82	−24.71	26.39	1.00	24.27	0.00	0.00	0.00
CA	0.010	0.48	−2.87	3.34	0.09	26.67	0.00	0.00	0.00		−0.10	4.30	−26.85	29.84	1.02	24.19	0.00	0.00	0.00
CO	0.011	0.51	−3.13	3.40	−0.23	25.14	0.00	0.00	0.00		−0.06	3.39	−21.64	23.62	1.18	24.71	0.00	0.00	0.00
CT	0.012	0.42	−2.47	2.90	0.25	25.67	0.00	0.00	0.00		−0.13	4.46	−28.46	27.80	0.57	22.17	0.00	0.00	0.00
DC	0.010	0.50	−3.03	3.17	−0.65	23.79	0.00	0.00	0.00		−0.11	3.83	–22.80	25.13	1.13	21.68	0.00	0.00	0.00
DE	0.010	0.47	−2.67	3.17	0.00	25.77	0.00	0.00	0.00		−0.13	4.30	−26.43	28.77	1.10	22.68	0.00	0.00	0.00
FL	0.011	0.45	−2.63	3.23	0.03	26.88	0.00	0.00	0.00		−0.06	3.57	–23.49	24.18	0.78	24.88	0.00	0.00	0.00
GA	0.011	0.47	−2.91	3.16	−0.42	27.46	0.00	0.00	0.00		−0.13	4.04	−24.34	28.92	1.39	24.33	0.00	0.00	0.00
HI	0.013	0.41	−2.44	2.85	0.13	26.75	0.00	0.00	0.00		−0.17	4.66	−28.56	31.53	0.80	22.16	0.00	0.00	0.00
IA	0.013	0.50	−2.83	4.02	0.59	27.98	0.00	0.00	0.00		−0.12	3.89	–22.81	37.56	3.62	44.83	0.00	0.00	0.00
ID	0.010	0.45	−2.50	3.04	0.34	20.67	0.00	0.00	0.00		−0.10	3.45	–23.13	21.85	0.51	23.14	0.00	0.00	0.00
IL	0.007	0.60	−4.02	3.80	−1.18	25.35	0.00	0.00	0.00		−0.05	3.40	−16.78	23.85	2.73	24.08	0.00	0.00	0.00
IN	0.012	0.43	−2.44	3.12	0.71	29.18	0.00	0.00	0.00		−0.09	3.88	−25.20	25.02	0.78	23.60	0.00	0.00	0.00
KS	0.012	0.43	−2.65	2.91	−0.27	28.24	0.00	0.00	0.00		−0.13	4.16	−26.05	26.95	0.62	22.92	0.00	0.00	0.00
KY	0.012	0.46	−3.01	2.78	−1.33	26.47	0.00	0.00	0.00		−0.10	3.97	–22.90	32.05	2.38	28.56	0.00	0.00	0.00
LA	0.012	0.43	−2.69	2.91	−0.66	27.17	0.00	0.00	0.00		−0.10	3.83	–23.37	26.02	1.36	24.34	0.00	0.00	0.00
MA	0.011	0.46	−2.66	3.23	0.37	27.51	0.00	0.00	0.00		−0.14	4.46	−26.88	30.33	1.00	22.94	0.00	0.00	0.00
MD	0.011	0.42	−2.49	3.05	0.58	28.80	0.00	0.00	0.00		−0.13	4.67	−28.29	32.84	0.99	24.49	0.00	0.00	0.00
ME	0.013	0.43	−2.58	3.30	0.83	30.53	0.00	0.00	0.00		−0.11	4.02	−26.21	25.32	0.36	22.36	0.00	0.00	0.00
MI	0.013	0.46	−2.66	3.55	0.44	29.46	0.00	0.00	0.00		−0.10	3.95	−24.42	27.60	1.22	24.63	0.00	0.00	0.00
MN	0.011	0.45	−2.81	3.07	−0.18	26.98	0.00	0.00	0.00		−0.10	3.86	−24.64	26.37	0.87	24.41	0.00	0.00	0.00
MO	0.012	0.49	−3.32	3.45	−0.25	29.68	0.00	0.00	0.00		−0.06	3.66	−24.09	23.74	0.77	24.32	0.00	0.00	0.00
MS	0.012	0.41	−2.44	2.92	−0.13	25.51	0.00	0.00	0.00		−0.12	3.90	−24.39	25.92	0.93	23.24	0.00	0.00	0.00
MT	0.017	0.50	−2.50	3.68	1.37	22.70	0.00	0.00	0.00		−0.07	3.84	−24.64	23.46	0.21	23.53	0.00	0.00	0.00
NC	0.012	0.42	−2.59	3.06	0.29	29.65	0.00	0.00	0.00		−0.15	4.44	−27.81	29.83	0.89	23.15	0.00	0.00	0.00
ND	0.012	0.56	−3.25	3.32	−0.43	20.05	0.00	0.00	0.00		0.00	3.66	–23.28	23.05	1.26	25.62	0.00	0.00	0.00
NE	0.012	0.45	−2.77	2.97	−0.79	25.92	0.00	0.00	0.00		−0.17	4.27	−25.44	26.52	1.03	19.74	0.00	0.00	0.00
NH	0.006	0.77	−4.25	5.08	0.14	23.28	0.00	0.00	0.00		−0.05	3.69	–22.88	22.67	0.99	21.16	0.00	0.00	0.00
NJ	0.014	0.55	−3.97	3.53	−2.49	29.68	0.00	0.00	0.00		−0.10	3.61	−18.29	29.04	2.81	28.05	0.00	0.00	0.00
NM	0.011	0.35	−2.35	2.20	−1.20	24.09	0.00	0.00	0.00		−0.18	4.26	−24.93	28.85	1.23	21.41	0.00	0.00	0.00
NV	0.011	0.47	−2.64	3.29	0.05	25.03	0.00	0.00	0.00		−0.09	4.05	−25.54	26.36	0.89	23.02	0.00	0.00	0.00
NY	0.009	0.46	−2.56	3.26	0.11	25.49	0.00	0.00	0.00		−0.05	4.26	−25.60	29.33	1.45	23.61	0.00	0.00	0.00
OH	0.012	0.48	−2.72	3.38	0.00	27.24	0.00	0.00	0.00		−0.05	4.39	−25.74	33.82	1.98	28.76	0.00	0.00	0.00
OK	0.011	0.44	−2.97	3.06	−0.70	27.20	0.00	0.00	0.00		−0.13	4.00	−25.11	25.77	0.50	23.84	0.00	0.00	0.00
OR	0.013	0.45	−2.68	3.24	0.50	28.62	0.00	0.00	0.00		−0.13	4.35	−28.44	27.54	0.36	22.90	0.00	0.00	0.00
PA	0.012	0.47	−2.58	3.41	0.40	27.83	0.00	0.00	0.00		−0.09	3.80	−24.58	26.57	0.89	25.31	0.00	0.00	0.00
RI	0.013	0.44	−2.70	3.18	0.31	28.83	0.00	0.00	0.00		−0.15	3.71	–23.68	23.72	0.50	22.68	0.00	0.00	0.00
SC	0.015	0.40	−2.41	2.72	−0.03	25.20	0.00	0.00	0.00		−0.13	4.05	–23.63	28.23	1.36	22.86	0.00	0.00	0.00
SD	0.015	0.51	−2.83	3.48	0.11	25.54	0.00	0.00	0.00		−0.11	3.72	−24.82	23.91	0.55	23.55	0.00	0.00	0.00
TN	0.012	0.42	−2.73	2.78	−0.52	26.85	0.00	0.00	0.00		−0.14	3.93	–23.68	27.86	1.28	23.81	0.00	0.00	0.00
TX	0.013	0.44	−2.61	3.10	0.14	27.51	0.00	0.00	0.00		−0.16	4.19	−26.75	28.15	0.78	24.03	0.00	0.00	0.00
UT	0.010	0.47	−2.87	3.14	−0.29	26.12	0.00	0.00	0.00		−0.13	4.51	−27.06	30.52	0.93	22.60	0.00	0.00	0.00
VA	0.013	0.43	−2.49	3.10	0.59	28.94	0.00	0.00	0.00		−0.15	4.21	−25.96	28.92	1.10	23.18	0.00	0.00	0.00
VT	0.012	0.44	−2.46	2.29	−1.07	16.48	0.00	0.00	0.00		−0.05	3.90	−24.75	25.63	0.67	23.65	0.00	0.00	0.00
WA	0.011	0.43	−2.45	2.89	−0.26	25.44	0.00	0.00	0.00		−0.15	4.49	−28.57	29.41	0.66	22.89	0.00	0.00	0.00
WI	0.012	0.38	−2.34	2.62	−0.01	27.44	0.00	0.00	0.00		−0.05	3.40	−21.01	23.33	1.19	25.00	0.00	0.00	0.00
WV	0.014	0.53	−3.40	3.71	−0.60	27.82	0.00	0.00	0.00		−0.07	3.84	−25.14	24.58	0.53	23.50	0.00	0.00	0.00
WY	0.016	0.59	−3.65	4.94	0.04	34.08	0.00	0.00	0.00		−0.10	3.36	−24.26	22.23	0.65	24.97	0.00	0.00	0.00

Panel B: Explanatory Variables
	Percent Change in 30-Day Average Positive Cases
COVID Covariate
State	Mean	Std.Dev	Min	Max	SK	KU	J-B (MSL)	DW (MSL)	ADF (MSL)
AK	1.98	5.10	−14.59	13.84	−0.48	1.11	0.00	0.00	0.05
AL	3.00	6.51	−7.62	27.23	1.83	3.50	0.00	0.00	0.12
AR	1.52	4.56	−11.68	13.00	−0.24	0.00	0.32	0.00	0.07
AZ	3.39	6.59	−12.97	22.65	0.60	0.86	0.00	0.00	0.20
CA	3.49	6.04	−7.13	24.80	1.50	2.56	0.00	0.00	0.24
CO	2.42	5.52	−6.35	19.65	1.12	1.27	0.00	0.00	0.23
CT	2.07	9.62	−15.45	29.19	0.77	0.25	0.00	0.00	0.18
DC	2.44	6.23	−11.12	24.70	1.49	2.78	0.00	0.00	0.12
DE	0.80	7.41	–22.44	20.22	0.20	0.95	0.01	0.00	0.12
FL	3.70	7.87	−7.53	32.48	1.69	3.04	0.00	0.00	0.13
GA	3.07	6.39	−7.68	25.94	1.63	2.74	0.00	0.00	0.22
HI	3.11	6.25	−11.30	21.28	0.38	−0.25	0.04	0.00	0.12
IA	2.38	5.97	−11.89	22.65	0.77	0.81	0.00	0.00	0.21
ID	0.76	6.03	−16.01	21.41	0.63	1.40	0.00	0.00	0.20
IL	3.88	9.06	−8.75	36.48	2.10	4.13	0.00	0.00	0.17
IN	2.45	5.33	−5.30	20.83	1.60	2.60	0.00	0.00	0.13
KS	3.29	4.88	−7.97	21.65	0.43	0.68	0.00	0.00	0.24
KY	1.30	6.93	−19.52	21.59	−0.45	1.58	0.00	0.00	0.11
LA	2.75	8.28	−14.98	28.64	0.95	0.65	0.00	0.00	0.18
MA	2.77	7.07	−13.47	25.43	1.29	2.32	0.00	0.00	0.17
MD	2.79	6.49	−8.75	29.48	1.92	3.98	0.00	0.00	0.14
ME	2.78	5.26	−12.36	18.97	0.49	0.72	0.00	0.00	0.12
MI	1.51	7.07	−18.96	21.85	0.96	1.30	0.00	0.00	0.09
MN	2.71	5.43	−12.66	16.49	0.47	−0.24	0.01	0.00	0.15
MO	2.14	6.73	−13.30	23.87	1.91	3.40	0.00	0.00	0.03
MS	2.64	6.76	−11.88	25.22	1.45	2.41	0.00	0.00	0.13
MT	2.19	6.36	−12.60	17.04	0.07	−0.29	0.56	0.00	0.07
NC	2.97	5.04	−7.32	20.89	1.55	2.79	0.00	0.00	0.17
ND	1.99	4.06	−13.06	13.10	−0.56	0.84	0.00	0.00	0.10
NE	2.51	5.46	−7.67	18.78	0.84	0.51	0.00	0.00	0.18
NH	3.08	6.51	−18.12	21.34	0.05	0.30	0.65	0.00	0.09
NJ	3.28	9.13	−11.05	37.21	2.03	4.09	0.00	0.00	0.13
NM	2.30	5.05	−7.83	18.38	0.70	0.40	0.00	0.00	0.19
NV	2.21	4.96	−8.92	18.44	0.72	0.64	0.00	0.00	0.13
NY	3.69	10.03	−9.21	38.93	2.15	4.41	0.00	0.00	0.10
OH	3.93	6.96	−4.55	27.35	2.09	4.01	0.00	0.00	0.11
OK	3.73	6.00	−8.62	23.07	1.51	2.21	0.00	0.00	0.16
OR	1.18	4.76	−12.44	15.66	0.15	0.66	0.10	0.00	0.05
PA	2.93	8.20	−15.51	31.37	1.60	3.11	0.00	0.00	0.16
RI	2.62	6.29	−8.40	22.88	0.73	−0.04	0.00	0.00	0.22
SC	2.75	4.85	−7.06	19.06	1.06	1.52	0.00	0.00	0.03
SD	1.75	5.20	−10.75	18.61	0.35	0.59	0.02	0.00	0.09
TN	2.76	4.41	−6.13	16.81	0.95	1.17	0.00	0.00	0.09
TX	3.22	6.03	−7.27	25.52	1.58	2.56	0.00	0.00	0.16
UT	2.94	5.38	−9.17	20.25	0.87	1.40	0.00	0.00	0.16
VA	2.91	6.94	−8.33	27.94	1.76	3.16	0.00	0.00	0.15
VT	1.84	5.74	−12.88	16.03	−0.23	−0.42	0.14	0.00	0.00
WA	3.25	7.56	−13.42	27.90	1.25	1.72	0.00	0.00	0.16
WI	3.23	6.51	−6.91	24.84	1.81	3.48	0.00	0.00	0.22
WV	2.91	5.36	−16.41	19.11	0.63	1.54	0.00	0.00	0.07
WY	1.03	4.56	−10.31	13.14	−0.25	−0.22	0.22	0.00	0.03
National	−0.855	20.37	−154.99	76.98	−2.30	17.23	0.00	0.00	1.00
Financial Covariates
RM	0.095	2.21	−12.77	8.97	−0.88	8.85	0.00	0.00	0.00
VIX	−0.080	8.78	−26.62	48.02	1.68	6.86	0.00	0.00	0.00
RMAA	0.012	0.44	−2.59	3.22	0.60	28.55	0.00	0.00	0.00
RMHY	0.010	0.78	−5.72	5.02	−1.76	27.85	0.00	0.00	0.00

Panel A presents the summary statistics of the municipal bond returns and bond yields while the same statistics for the COVID-19 variable and the other four financial covariates are reported in Panel B.

Concerning the measures of the prevalence rate of COVID-19 (which is the 30-day moving average of the daily changes in the number of positive cases), it is evident that there are variations in the prevalence rates of the virus across U.S. states. Lower values of the average daily changes were obtained for low-density states such as Idaho, Delaware, Wyoming, and Oregon. While the highest values of the average daily changes were obtained for densely populated states such as Ohio, Illinois, Florida, New York, and California. The skewness and kurtosis statistics, as well as the Jarque-Bera, Durbin-Watson, and the Augmented Dickey fuller tests, suggest that the COVID prevalence variable for the 46 states and the District of Columbia (excluding Arkansas, New Hampshire, Vermont, Wyoming, and Montana) are also not normally distributed.

Turning our attention to the results (Panel B) which were obtained for the financial covariate variables (S&P 500 (RM), the Bloomberg Barclays AAA Municipal Bond Index (RMAA), the Bloomberg Barclays High Yield Municipal Bond Index (RMHY), and the daily changes in the value of the CBOE Volatility (VIX) index (RVIX)). The average return on the S&P 500 is higher than those of municipal bonds in general (including high-quality and low-quality municipal bonds). The standard deviation of the S&P 500 is also greater than the same values for all state municipal bonds, thus supporting the notion that equities hold higher risk than bonds. The average daily returns for both the high-yield municipal bond (0.010%) and the AAA municipal bond indices (0.012%) are in line with the return of the bond indices for the states. On the contrary, the standard deviation of the return of the high yield municipal bond index (0.78%) is greater than those obtained for the state indices, while the same statistic for the AAA municipal bond index (0.44%) is more in line with the state indices.

3. Accounting for the aggregate effects of the COVID-19 pandemic on the municipal bond Market.

In this section, we scrutinize the aggregate impact of the COVID-19 pandemic on the municipal bond market using various econometric specifications. First, we construct a panel regression of the state-level municipal bonds returns on their respective COVID-19 prevalence rates. Consider a standard panel regression of the form:

$$R_{\mathit{it}}=\alpha +X_{\mathit{it}}^{\prime }\beta +u_{\mathit{it}}$$ 1

Where $X_{\mathit{it}}$ is a set of explanatory variables such that:

$X_{\mathit{it}}=\left[R{M}_t,RVI{X}_t,RMA{A}_t,RMH{Y}_t,COVID{R}_{\mathit{it}}\right]$ and $\beta$ is a $K\times 1$ vector of coefficients. $R{M}_t$ is the daily percent change in the S&P 500 index, $\mathit{RVI}{X}_t$ is the daily percent change in the VIX index, $\mathit{RMA}{A}_t$is the daily percent change in the Bloomberg Barclays AAA municipal bond index, $\mathit{RMH}{Y}_t$is the daily percent change in the Bloomberg Barclays High Yield municipal bond index, and $\mathit{COVID}{R}_{\mathit{it}}$ is the proxy for COVID-19 prevalence rates in each state. Let $R_{\mathit{it}}$ and $X_{\mathit{it}}$ be matrices formed by arranging the municipal bond returns and the explanatory variables by state and by time across each state. 15 Augmenting the first column $X_{\mathit{it}}$ with a vector of ones, which corresponds to the intercept term $\alpha$ yields the general framework for the 5-panel regression models shown below:

Pooled OLS:

$$u_{it}={\epsilon }_{it}$$ 2

Where ${\epsilon }_{\mathit{it}}$ are iid with zero mean and variance${\sigma }_{\epsilon }^2$. In the pooled OLS, all observations are pooled and estimated using simple OLS.16

Random Effect:

$$u_{\mathit{it}}={\nu }_i+{\epsilon }_{\mathit{it}}$$ 3

Where ${\nu }_i$ are iid with zero mean and variance ${\sigma }_{\nu }^2$, and ${\epsilon }_{\mathit{it}}$ are iid with zero mean and variance ${\sigma }_{\epsilon }^2$. The estimation of the Random effect specification is made under the assumption that the independent variables are uncorrelated with the unobservable state-specific effects and observation-level errors.

Given that a set of covariates are cross-sectionally invariant, we also include between-time period pooled regression, where the daily averages of the municipal bond returns and yield are regressed against the average daily changes in the COVID-19 prevalence rates and the financial covariates as shown below.

Between Time Periods:

$${\overline{R}}_t=\alpha +{\overline{X}}_t^{\prime }\beta +{\overline{\epsilon }}_t$$ 4

Where ${\overline{R}}_t$ is the average daily percent changes in municipal bond returns, and ${\overline{X}}_t$ is the matrix of explanatory variables that have been collapsed into daily time period means.

A remarkable aspect of the COVID-19 pandemic is that almost every sphere of human activity was impacted in some way by it. Thus, we also address the possibility that individual state-specific unobservable effects could be correlated with other explanatory variables by estimating the Hausman-Taylor (1981) model and the System Generalized Method of Moments (SGMM) using the econometric specifications below.

Hausman Taylor:

$$R_{\mathit{it}}=\alpha +X_{\mathit{it}}^{\prime }\beta +Z_i^{\prime }\gamma +{\nu }_i+{\epsilon }_{\mathit{it}}$$ 5

Where $Z_i$ are cross-sectional time-invariant variables $\left[Z_1,Z_2\right]$ and the time-varying regressors in $X_{\mathit{it}}$are split into two sets $X_{1it}$ and$X_{2it}$. $X_{1it}$denotes the exogenous variables that are independent of both error terms (namely the state-by-state COVID-19 prevalence rates, $\mathit{COVID}{R}_{\mathit{it}}$), while $X_{2it}=\left[R{M}_t,RVI{X}_t,RMA{A}_t,RMH{Y}_t\right]$denotes variables that are independent of the observation-level errors ${\epsilon }_{\mathit{it}}$ but may be correlated with state-specific errors ${\nu }_i$. The Hausman and Taylor (1981) model is a two-stage least square (2SLS) hybrid regression that combines the consistency of a fixed-effects model with the efficiency and applicability of a random-effects model. The instruments ${\stackrel{\sim }{X}}_{1it}$ and ${\stackrel{\sim }{X}}_{2it}$are the “within” transformation which represents the deviation of each value of the time-varying regressors from its cross-sectional means.

System Generalized Method of Moments (SGMM):

$$R_{\mathit{it}}=\alpha +{\mathrm{\Pi }}_{\mathit{it}}^{\prime }\beta +W_t^{\prime }\gamma +{\nu }_i+{\epsilon }_{\mathit{it}}$$ 6

Where ${\mathrm{\Pi }}_{\mathit{it}}$ is a set of strictly exogenous regressors (namely the state-by-state COVID prevalence rates, $\mathit{COVID}{R}_{\mathit{it}}$) and $W_t$is a set of endogenous regressors such that$W_t=\left[R{M}_t,RVI{X}_t,RMA{A}_t,RMH{Y}_t\right]$. To remove the source of the correlation, we could attempt to take the first difference of Equation (6), which removes ${\nu }_i$ and its omitted variable bias as shown below.

$$\mathrm{\Delta }{R}_{\mathit{it}}=\alpha +\mathrm{\Delta }{\mathrm{\Pi }}_{\mathit{it}}\beta +\mathrm{\Delta }{W}_t\gamma +{\eta }_{\mathit{it}}$$ 7

However, the problem of endogeneity still persists because $W_t$ may be correlated with the error term ${\eta }_{\mathit{it}}$. Thereby motivating the need for a set of instruments that can fulfill the orthogonality conditions$\mathrm{E}\left[{\mathrm{Z}}_{\mathit{it}},{\eta }_{\mathit{it}}\right]=0$. In the system GMM approach, the first difference specification shown in Equation (7) is augmented by stacking Equation (7) on top of the level specification in Equation (6), such that the system instrument matrix $Z_i^s$ is now a stacking of the lagged first differences of the$\mathrm{\Delta }{R}_{\mathit{it}-1}$ and$R_{\mathit{it}-1}$ according to the specification below.

$$Z_i^s=\left(\begin{array}{cc}{Z}_i^d& 0\\ 0& Z_i^l\end{array}\right)$$ 8

Where $Z_i^d=\left(\begin{array}{cccccccccc}{R}_{\mathit{it}-1}& 0& 0& 0& 0& 0& \cdots & 0& 0& 0\\ 0& R_{\mathit{it}-1}& R_{\mathit{it}-2}& 0& 0& 0& \cdots & 0& 0& 0\\ 0& 0& 0& R_{\mathit{it}-1}& R_{\mathit{it}-2}& R_{\mathit{it}-3}& 0& \cdots & 0& 0\\ ⋮& ⋮& ⋮& ⋮& ⋮& ⋮& \ddots & ⋮& ⋮& ⋮\\ 0& 0& 0& 0& 0& 0& 0& R_{\mathit{it}-1}& \cdots & R_{i,T-2}\end{array}\right)and{Z}_i^l=\left(\begin{array}{ccccc}0& 0& 0& \cdots & 0\\ 0& \mathrm{\Delta }{R}_{it-1}& 0& \cdots & 0\\ 0& \mathrm{\Delta }{R}_{it-2}& \cdots & 0& \\ ⋮& ⋮& ⋮& \ddots & 0\\ 0& 0& 0& \cdots & \mathrm{\Delta }{R}_{i,T-1}\end{array}\right)$

Table 2 reports the results of the panel regression using the specifications described above, where the 30-Day moving average of the percent increase in COVID-19 cases was employed as the proxy for the COVID-19 effect. In Panel A, we report the COVID-19 effect on aggregate performances of municipal bond returns. The results of the same regression specification using municipal bond yields are reported in Panel B. On the whole, the results show that COVID-19 prevalence is a risk factor that market participants price into the performances of municipal bonds. This assertion is supported by the statistically significant negative coefficients which were obtained for the COVID-19 proxies for the municipal bond returns and the statistically significant positive coefficients for the municipal bond yields in the panel regressions. The findings above are robust to the inclusion of other explanatory variables, including the daily percent changes in the Bloomberg Barclays AAA municipal bond index and the daily percent changes in the Bloomberg Barclays High Yield Municipal Bond index. Going by the coefficients of the $\mathit{RMA}{A}_t$ variable in both Panels A and B, it is evident that muni bond indices are mostly comprised of high-quality municipal bonds.17 The bond yield regressions also highlight that yields on municipal bonds are significantly correlated with the levels of risk in the equity market as shown by the values of the $\mathit{RVI}{X}_t$ variable. Municipal bond yields are also negatively correlated with price levels in the general municipal bond markets, with a higher sensitivity towards the high-quality municipal bond market.

Table 2.

Panel Regressions of COVID-19 Effect on Municipal Bond Performances.

	Panel A Municipal Bond Returns									Panel B Municipal Bond Yields
	Constant	RM	RVIX	RMAA	RMHY	COVIDR	R-Sq			Constant	RM	RVIX	RMAA	RMHY	COVIDR	R-Sq
RDM-Effect	0.005	0.004	0.001	0.934	0.052	−0.002	0.922			−0.043	0.061	0.030	−7.315	−0.863	0.012	0.913
	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)				(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)
Hausan Taylor	0.005	0.004	0.001	0.934	0.052	−0.002	0.922			−0.043	0.061	0.030	−7.315	−0.863	0.012	0.913
	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)				(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)
Pooled	0.005	0.004	0.001	0.934	0.052	−0.002	0.922			−0.043	0.061	0.030	−7.315	−0.863	0.012	0.913
	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)				(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)
Between Time Periods	0.008	0.004	0.001	0.948	0.044	−0.003	0.994			−0.064	0.061	0.031	−7.392	−0.823	0.020	0.945
	(0.00)	(0.02)	(0.00)	(0.00)	(0.00)	(0.00)				(0.36)	(0.16)	(0.00)	(0.00)	(0.00)	(0.15)
GMM	0.008	0.003	0.001	0.998	0.021	−0.004				−0.077	0.064	0.033	−7.648	−0.643	0.025
	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)				(0.00)	(0.00)	(0.00)	(0.00)	(0.00)	(0.00)

Panels A and B present results pertaining to municipal bond returns and municipal bond yields, respectively. p-values are in parentheses. All significant coefficients at the 10% level or below are in bold. RM is the market return, defined as the percentage change in the S&P 500 Index, RVIX is the daily changes in the value of the CBOE Volatility (VIX) Index, RMAA is the Bloomberg Barclays AAA Municipal Bond Index, RMHY is the Bloomberg Barclays High Yield Municipal Bond Index and COVIDR is the rate of prevalence of COVID-19.

Fig. 1 graphs the time series of the S&P U.S. Municipal Bond yield index and the 30-Day moving average of the percent changes in U.S. COVID-19 cases. Evident on the graph is a plot that suggests that the aggregate performances of U.S. municipal bonds are negatively correlated with the prevalence rate of the virus. Particularly in the early periods of the pandemic, where the average yield of U.S. municipal bonds spiked by over 160 basis points (from approximately 4.10% to over 5.77%). Further evidence in support of this relationship is also shown in Fig. 2 , which graphs the Bloomberg Barclays U.S. Municipal Bond index and the 30-Day moving average of the percent changes in U.S. COVID-19 cases. Apparent on the graph is a crash in the value of U.S. municipal bonds at the advent of the crisis. This was followed by a subsequent recovery in the value of the bonds as the large percentage increases in cases that were observed at the beginning of the pandemic subsided over time.

Fig. 1.

U.S COVID-19 Prevalence and U.S Municipal Bond Yields.

Fig. 2.

U.S COVID-19 Prevalence and U.S Municipal Bond Index Levels.

4. COVID-19 and State-Level municipal bond Performances.

Given the disparities in the policy response to the pandemic and the variation in the economic fundamentals across U.S. states, it is worthwhile to conduct a state-by-state assessment of the impact of the pandemic on municipal bond performances. First, we regressed each state’s municipal bond index returns and municipal yield index on its COVID-19 prevalence rate using the econometric specification below.

$$R_{it}=\alpha +{\beta }_{1}R{M}_t+{\beta }_{2}RVI{X}_t+{\beta }_{3}RMA{A}_t+{\beta }_{4}RMH{Y}_t+{\beta }_{5}COVID{R}_{it}+{\epsilon }_{it}$$ 9

In Table 3 , we report the results of the regression specification shown above, where the 30-Day moving average of the percent change in COVID-19 cases was employed as the proxy for the COVID-19 effect. In Panel A, we report the coefficients of the COVID-19 variable for all 50 states and the District of Columbia. The same regression coefficients for each state’s municipal bond yields are reported in Panel B.18 Overall, the results show COVID-19 effects are state-specific. The coefficient of the variable which accounts for the effect of the prevalence of COVID-19 in each state on its municipal bond return was negative and significant in 23 states at the 5% level of significance and 4 states (including the District of Columbia) at the 10% level of significance. The same specification, which examines the impact on municipal bond yields was positive and significant in 14 states at the 5% level and in 3 states (including the District of Columbia) at the 10% level.

Table 3.

COVID Effects on State-Level Municipal Bond Performances.

	Panel A Bond Returns		Panel B Bond Yields
State	Coefficient	P-Value	Coefficient	P-Value	Group
AL	0.001	(0.44)	−0.001	(0.91)	2
AK	0.000	(0.71)	0.004	(0.64)	2
AZ	−0.001	(0.11)	0.006	(0.28)	2
AR	−0.001	(0.57)	0.010	(0.04)	2
CA	−0.001	(0.11)	0.026	(0.00)	2
CO	−0.002	(0.01)	0.024	(0.00)	1
CT	−0.001	(0.06)	0.010	(0.03)	1
DE	0.000	(0.51)	−0.005	(0.39)	2
DC	−0.002	(0.06)	0.009	(0.09)	1
FL	−0.001	(0.00)	0.008	(0.09)	1
GA	−0.002	(0.00)	0.005	(0.45)	1
HI	0.001	(0.25)	−0.002	(0.78)	2
ID	−0.006	(0.00)	−0.005	(0.57)	1
IL	−0.008	(0.00)	0.011	(0.01)	1
IN	−0.003	(0.00)	0.046	(0.00)	1
IA	0.000	(1.00)	0.003	(0.64)	2
KS	−0.002	(0.07)	0.002	(0.77)	1
KY	−0.001	(0.42)	0.004	(0.53)	2
LA	−0.002	(0.00)	0.000	(1.00)	1
ME	0.000	(0.60)	0.010	(0.13)	2
MD	−0.001	(0.03)	−0.009	(0.27)	1
MA	−0.001	(0.06)	0.003	(0.66)	1
MI	−0.001	(0.03)	0.021	(0.00)	1
MN	−0.002	(0.00)	0.027	(0.00)	1
MS	−0.001	(0.09)	−0.002	(0.62)	1
MO	−0.003	(0.00)	−0.005	(0.30)	1
MT	0.000	(0.97)	0.006	(0.11)	2
NE	0.001	(0.37)	0.000	(0.99)	2
NV	0.000	(0.92)	0.003	(0.70)	2
NH	−0.009	(0.01)	0.011	(0.15)	1
NJ	−0.006	(0.00)	0.005	(0.38)	1
NM	0.001	(0.64)	0.019	(0.01)	2
NY	−0.002	(0.00)	0.019	(0.04)	1
NC	−0.001	(0.01)	0.001	(0.93)	1
ND	0.000	(0.98)	0.022	(0.00)	2
OH	−0.002	(0.00)	0.001	(0.87)	1
OK	−0.001	(0.26)	−0.001	(0.90)	2
OR	0.000	(0.67)	0.008	(0.21)	2
PA	−0.002	(0.00)	0.010	(0.05)	1
RI	−0.001	(0.32)	0.001	(0.89)	2
SC	−0.005	(0.00)	−0.011	(0.11)	1
SD	0.000	(0.80)	0.004	(0.46)	2
TN	0.000	(0.77)	0.000	(0.96)	2
TX	−0.001	(0.00)	0.007	(0.23)	1
UT	0.000	(0.98)	0.004	(0.59)	2
VT	0.000	(0.84)	0.032	(0.00)	2
VA	−0.002	(0.00)	0.004	(0.27)	1
WA	−0.001	(0.03)	0.004	(0.52)	1
WV	−0.002	(0.25)	0.018	(0.00)	2
WI	−0.002	(0.00)	0.001	(0.77)	1
WY	0.003	(0.25)	0.012	(0.01)	2

In the next step, we explore whether the economic fundamentals and/or the COVID-19 policies of each state can explain the sensitivity of the state’s municipal bonds to the COVID-19 prevalence rates. For economic fundamentals, we examined whether the disparities in the state and local tax collections per capita (STC), state and local revenue per capita (SRC), federal aid as a percent of general revenue (FTA), total tax burden per capita (TPC), population density (PDC), debt per capita (DPC), and income per capita (IPC) can explain the COVID-19 sensitivities of the state’s municipal bond performances.19 Furthermore, we also examined whether the implementation of the following COVID-19 mitigation policies by each state had a discriminating role in explaining the COVID-19 sensitivities of the municipal bond issued in that state, namely: stay-at-home mandate (SAHDUR), quarantine mandate (Q-MAN), face mask mandate (FMM), closure of K-12 public schools (CK12), closure of daycare centers (CDCC), banned visitors to nursing homes (CNH), closure of other non-essential businesses (CNEB), closure of restaurants (CRB), closure of gyms (CGB), closure of movie theaters (CMTB), closure of bars (CBB), closure of casinos (CCB), restriction on overnight business services (CBO).20 Classifying the policy response from each state was challenging, due to the inconsistent application of the policies at the different levels of government. Indeed, some states implemented various policies at the onset of the pandemic, only to rescind them after a very short period of time. Given the duration of the study, it is unlikely that mitigation policies that did not remain in place for at least 30 days will be efficacious in mitigating the pandemic in the states. Thus, a state was classified as enacting a mitigating policy, only if the policy remained in place for at least a month.

In Table 4 , we group all states and the District of Columbia into those with significant COVID effects on the municipal bond returns and those without. Subsequently, we report the summary statistics for the economic fundamentals and policy responses for each group. Furthermore, we assess whether the economic and policy response characteristics of each group can explain the disparities in the COVID-19 effect using the tests for the equality of means for economic fundamentals and the test for the equality of proportions for the policy response. Overall, states with COVID-19 effects appear to be more reliant on tax receipts, are more densely populated, have higher degrees of indebtedness, and have higher per capita income as the test of means for these two groups show statistically significant differences. Along the same line, the COVID-19 sensitive states also appear to have implemented more COVID-19 mitigating policies than those without. For example, about 81.5% and 92.6% of the states with the COVID-19 effect enacted a stay-at-home and face mask mandate, respectively, compared to 50% and 75% of the states without the COVID-19 effect.

Table 4.

Economic Fundamentals, Policy Responses, and COVID-19 Effects.

Panel A: Economic Fundamentals
								STC	SRC	FTA	TPC	PDC	DPC	IPC	SEF
States with COVID Effect	Mean							$5,526	$10,078	31%	$5,846	605	$9,505	$57,365	164
	Std.Dev							$1,909	$2,580	6%	$1,749	1,909	$4,249	$10,986	72
	Min							$3,705	$7,241	20%	$3,654	22	$3,494	$38,914	298
	Max							$11,311	$19,169	45%	$9,987	10,083	$23,031	$83,406	29
States without COVID Effect	Mean							$4,944	$10,325	34%	$4,982	102	$8,014	$51,928	194
	Std.Dev							$1,171	$2,327	6%	$1,132	157	$2,519	$6,501	57
	Min							$3,286	$6,983	21%	$3,368	1	$3,325	$42,315	290
	Max							$7,611	$15,630	44%	$7,529	711	$12,553	$66,619	72
Test of Means	COVID > NOCOVID							(0.10)	(0.36)	(0.06)	(0.02)	(0.09)	(0.06)	(0.02)	(0.06)
	COVID = NOCOVID							(0.19)	(0.72)	(0.11)	(0.04)	(0.18)	(0.13)	(0.03)	(0.11)
Panel B: Policy Responses
COVID Response Policies and COVID Effect on Municipal Bond Returns
			SAHDUR	Q-MAN	FMM	CK12	CDCC	CNH	CNEB	CRB	CGB	CMTB	CBB	CCB	CBO
States with COVID Effect			81.5%	44.4%	92.6%	100.0%	25.9%	70.4%	85.2%	96.3%	96.3%	92.6%	100.0%	48.1%	44.4%
States without COVID Effect			50.0%	62.5%	75.0%	95.8%	33.3%	50.0%	83.3%	91.7%	91.7%	91.7%	95.8%	37.5%	37.5%
Test of Proportions	COVID > NOCOVID		(0.01)	(0.90)	(0.04)	(0.14)	(072)	(0.07)	(043)	(0.24)	(0.24)	(0.45)	(0.14)	(0.22)	(031)
	COVID = NOCOVID		(0.02)	(0.20)	(0.08)	(028)	(0.56)	(0.14)	(0.21)	(048)	(0.48)	(0.90)	(0.28)	(0.44)	(0.61)
COVID-19 Response Policy Index
			Number of Mitigation Policies Implemented							Proportion of Mitigation Policies Implemented
					Mean	Max	Min				Mean	Max	Min
COVID-19 Effect					10	13	4				75%	100%	31%
No COVID-19 Effect					9	12	1				69%	92%	0%

This table presents the results of all states and the District of Columbia grouped into those with significant COVID effects on the municipal bond return and those without. The economic fundamental variables are defined as follows: state and local tax collections per capita (STC), state and local revenue per capita (SRC), federal aid as a percent of general revenue (FTA), total tax burden per capita (TPC), population density (PDC), debt per capita (DPC), income per capita (IPC), and state economic fundamentals (SEF). The COVID-19 response variables are defined as: stay-at-home mandate (SAHDUR), quarantine mandate (Q-MAN), face mask mandate (FMM), closure of K-12 public schools (CK12), closure of daycare centers (CDCC), banned visitors to nursing homes (CNH), closure of other non-essential businesses (CNEB), closure of restaurants (CRB), closure of gyms (CGB), closure of movie theaters (CMTB), closure of bars (CBB), closure of casinos (CCB) and restriction on overnight business services (CBO). In Panels A and B, the values in parentheses are the p-values of the test of the equality of means (Panel A) and the equality of proportions (Panel B).

A summary of the unique positioning of each state is accounted for through the state economic fundamentals (SEF) and policy response indices. The state economic fundamentals index is a sum of the Tax Foundation’s desirability ranking (1 is best and 50 is worst) of each state based on the 6 fundamental measures (STC, SRC, FTA, TPC, PDC, DPC, IPC) and population density score (1 is lowest and 50 is highest).21 In line with the ranking methodology described above, the lower the overall economic fundamentals index of a state, the higher the economic desirability of the state. The policy response index is also calculated as the total number of mitigation policies implemented by each state. Therefore, the greater the number of policy responses implemented, the higher the policy response index score for the state. The mean economic fundamentals index (1 6 4) for states with COVID-19 effects is significantly lower than for states without COVID-19 effects (1 9 4). This implies that states with COVID-19 effects appear to have more favorable economic fundamentals than states without. In contrast, the average number of policy responses from states with (10 responses), and without (9 responses) the COVID-19 effect appears to be much closer, albeit with a wider range which is evident in the latter group. Overall, the above suggests that economic fundamentals may be the primary driver of the sensitivities of each states’ municipal bond performance to its COVID-19 prevalence rate.

5. Predicting COVID-19 sensitivity of State-Level municipal bond Performances.

In this section, we explore whether the interaction of the economic fundamentals and policy responses can shed more light on the COVID-19 sensitivities of each state's municipal bond return to its COVID-19 prevalence rates. In particular, we explore whether a multivariate framework of fundamental factors and policy responses can discriminate between states with COVID-19 response and the states without. First, we create a cross-sectional dichotomous variable that represents the grouping of states based on the COVID-19 sensitivity of its municipal bond returns. Next, we implement both linear and canonical discriminant analysis to assess whether some linear combination of the economic fundamentals factors and policy responses can predict the COVID-19 sensitivity of the state’s municipal bond performance. In the linear discriminant analysis framework, a value score is derived for each observation such that when compared with the class boundaries derived by the discriminant function and the statistical properties of the predictors, a prediction for the class of the observation is derived. Consider a value function $G(S_j,k)$, which measures the likelihood of state $S_j$ belonging to class $k=2$ (COVID EFFECT and NOCOVID-EFFECT), thus the discriminant score $D_j$ for assigning $S_j$ to one of the two classes is given by.

$$D_j=G(S_j,k_1)-G(S_j,k_2)$$ 10

Hence state $S_j$ is assigned to the COVID-EFFECT class if $D_j$ is positive, and to the NOCOVID-Effect class if $D_j$ is negative. The discriminant score $D_j$is calculated from the linear combination of the values of the 8 predictors variables, which include the 7 economic fundamentals scores (STC, SRC, FTA, TPC, PDC, DPC, IPC) and the policy response index (CPR) for each state as shown below.

$$D_j=\sum _j^N{\mathrm{\Pi }}_{\zeta }{\mathrm{\Phi }}_j$$ 11

Where ${\mathrm{\Pi }}_{\zeta }$ is a vector of discriminant parameters and ${\mathrm{\Phi }}_j$is the matrix of the predictor variables described above.

Table 5 reports the result of the model specifications in Equations 10, 11. In Panel A, the estimated linear discriminant functions are reported. In Panel B, canonical variables which represent a set of the linear combination of the predictor variables that best reveals the difference between the classes are also reported. Canonical variables were obtained from the canonical discriminant analysis, which is a dimension reduction model that is similar to the principal component analysis.22 The multivariate test of the predictive ability of the discriminant function and the canonical variables are reported in Panel C. Going by the p-values of the multivariate tests, it is evident that linear combinations of the values of the 8 predictors variables have a superior ability to determine the sensitivities of the municipal bond performances of each state to COVID-19. The values obtained from the linear discriminant functions suggest that states with higher dependences on tax receipts (STC), lower revenue per capita (SRC), lower reliance on federal spending (FTA), lower tax burden per capita (TPC), higher population density (PDC), higher debt per capita (DPC), higher income per capita (IPC), and implemented more mitigation policies are more likely to have municipal bond performances with sensitivities to COVID-19 prevalence rates.23 Further evidence of this can be found in the canonical coefficients of the canonical variable 1 which shows positive relationships between the COVID-19 sensitivities of municipal bond returns and STC, PDC, DPC, IPC, and the COVID-19 policy response index (CPR), while showing negative correlations between COVID-19 sensitivities and SRC, FTA, and TPC.

Table 5.

The Determinants of State-Level Municipal Bond Response to COVID-19.

Panel A: Linear Discriminant Function				Panel B: Canonical Variables
Variable	COVID Effect		NOCOVID Effect		Canonical Variable 1		Canonical Variable 2
Constant	−73.61		−69.46
STC	7.07		7.01		0.04		−1.06
SRC	−3.73		−2.85		−0.66		0.32
FTA	169.80		171.31		−1.14		6.91
TPC	−5.57		−5.40		−0.13		1.84
PDC	−2.47		−2.85		0.29		0.54
DPC	−1.38		−1.44		0.05		−0.25
IPC	2.14		1.94		0.15		−0.08
CPR	21.53		17.70		2.89		−2.34
Panel C: Test Statistics
	Multivariate Test
Statistic	Value	P-Values		Canonical Correlation			0.5597
Wilks' Lambda	0.687	(0.03)		Adjusted Canonical Correlation			0.4777
Pillai's Trace	0.313	(0.03)		Squared Canonical Correlation			0.3133
Hotelling-Lawley Trace	0.456	(0.03)		P-value			(0.03)
Roy's Greatest Root	0.456	(0.03)		Error Rate			21.57%

The results of the discriminant analysis are presented in this table. Panel A presents the estimated linear discriminant functions while the canonical variables are reported in Panel B. The economic fundamental variables are defined as follows: state and local tax collections per capita (STC), state and local revenue per capita (SRC), federal aid as a percent of general revenue (FTA), total tax burden per capita (TPC), population density (PDC), debt per capita (DPC), income per capita (IPC), and policy response index (CPR).

The ability of the linear discriminant model to correctly classify the COVID sensitivities of each state based on its economic fundamentals and policy response is buttressed by the canonical correlation which shows that 56% of the within the group variation can be explained by the model. Furthermore, the error rate reported in Table 5 indicates that the model accurately predicts 78% of the cases correctly. Fig. 3 depicts the prediction ability of the model. On the graph, the distribution of the canonical variables 1 and 2 are overlayed with the model prediction for each state. Apparent on the graph is the ability of canonical variable 1 to distinguish states with COVID effect, from states without COVID effect, while the canonical variable 2 is able to do otherwise. Taken together, the empirical results obtained from this section elucidate the roles of economic fundamentals in the COVID sensitivities of municipal bond performances. Strong economic fundamentals imply a higher level of economic activities, which makes the state’s finances more susceptible to the decline in economic activities that arose from the pandemic.

Fig. 3.

Predicting State Municipal Bond Sensitivity to COVID-19 Effect.

6. Conclusion.

The COVID-19 pandemic has left a devastating impact on people, businesses, and governments around the world. As of the time of writing this paper, the economic and human toll of the pandemic is astounding, with over 600 thousand deaths and over 30 million cases in the U.S. alone. During the month of April 2020, the U.S. experienced the largest job loss in history, with over 20 million jobs lost in a single month. In the year 2020, the U.S. government also ran the largest post-war deficit ($3.1 trillion) as a share of GDP (15.2%) in history. The effects of the pandemic were also apparent on the finances of U.S. states and municipalities as total tax receipts fell by more than 2.5%, while total expenditure increased significantly over prior years.24 In this study, we attempt to quantify the economic effect of the virus on the financial health of U.S. states and municipalities as measured by the performance of the municipal bonds issued in each state. In particular, the study analyzes the impact of the prevalence rates of the COVID-19 virus across the U.S. states and its associated effect on the aggregate performances of the municipal bonds issued within each state. Using panel regressions and state-by-state regressions, the study finds significant evidence that prevalence levels of the COVID-19 virus negatively impacted the aggregate performances of the municipal bonds issued in some states. In general, municipal bond returns (yields) were found to be negatively (positively) affected by COVID-19 prevalence rates. The result is robust after controlling for other determinants of municipal bond performances, including the stock market performance, and the aggregate performance of investment quality and high-yield municipal bonds. Given the disparities in the human and economic cost of the virus across the 50 states and the District of Columbia, and in the policy responses enacted by states and local governments to mitigate the impact of the virus, the study also explored whether the economic fundamentals and the COVID-19 mitigation policies enacted by each state can explain the sensitivities of the aggregate performances of municipal bond outstanding to the COVID-19 prevalence rates. We find that states with higher dependences on tax receipts, lower revenue per capita, lower reliance on federal spending, lower tax burden per capita, higher population density, higher debt per capita, higher income per capita, and implemented more mitigation policies are more likely to have municipal bonds that were negatively impacted by the COVID-19 prevalence rates. A possible explanation for this finding is that municipal bonds’ performances are channels through which investors account for the economic risk of the mitigation policies, which were implemented to prevent the further spread of the virus within each state. In most states, the mitigation policies led to a substantial decline in economic activities and precipitated massive job and income losses, which then led to a substantial decline in government receipts. Since municipal bonds are obligations that are funded by the expected future cash flows to the issuer, the negative effect observed may be an upshot of investors’ expectation that the decline in government receipts would lead to a deterioration in the credit qualities of bonds issued by entities in that state.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

The authors do not have permission to share data.