The $3.8 trillion U.S. municipal bond market has been firmly rooted in the country’s capital markets since the first recorded general obligation issued by the City of New York in 1812. In contrast to equities and corporate bonds where quarterly earnings and M&A activity often take center stage, municipal bond investors focus more on political dynamics, public policy, economic data, tax collections, as well as outstanding pension liabilities and other postemployment benefits (OPEB). For example, it is not uncommon for a municipal bond issue to report audited financials only annually, as the audit frequency and breadth of disclosures in the report are often driven by legislation and difficult to change.
As a result, municipal bond investors are beginning to examine more “non-traditional” datasets for making investment decisions and surveillance given the potential for wide gaps between reporting periods. A broad assortment of alternative data has already been mined for alpha signals and incorporated into equity trading algorithms for several years, while its application for the municipal bond market is only in its infancy in comparison. We believe that the inevitable rising rate environment will take away some of the cushion that has recently allowed municipalities to refinance into lower debt payments, which will increase the demand for new sources of data to more effectively price and tier municipal bond risk.
In late-2012, Puerto Rico’s deteriorating economic conditions led to concerns over the Commonwealth’s ability to service its approximately $72 billion in outstanding municipal debt and the island’s various bond issues’ prices dropped based on the income sources and covenants of each issuing entity. It wasn’t until late-June 2015 that the government announced that its current debt burden was “unpayable,” which made a municipal bond default appear inevitable. Since filing for Title III protection on May 3, Puerto Rico’s benchmark General Obligation 8.0% 7/2035 hit a new record low closing price of 58.25 on May 23 (Figure 1) and ended July only slightly higher at a price of 59.50.
Investors in the island’s debt continue look for new approaches to assessing and projecting its economic condition. According to data from the Government Development Bank for Puerto Rico, the services industry made up 12.9% of the island’s GDP in 2015 (the last year the island reported). Additionally, almost 35% of the island was employed by the services industry in 2016, with a large portion of them working in tourism and hospitality. Given that approximately a third of tourists enter the island on cruise ships, their traffic could provide an indicator of the health of that industry on the island.
We recently measured cruise ship traffic to monitor the daily number of cruise ships at port. Using almost real time ship tracking technology from IHS Markit, we tracked the number of cruise ships at port in Puerto Rico from 2014 until June 2017. This is similar to the methods that some quantitative investment managers’ use of lagged satellite imagery of large retailer’s parking lots in an attempt to project store revenue.
The cruise ship data did correlate to some degree with general fund gross revenues (includes most taxes, federal revenue, and certain miscellaneous revenue) reported by Puerto Rico’s Departamento de Hacienda (Figure 2). The correlation is most apparent when the cruise ship data is lagged four months, as the lag overlays the peak cruise ship month of December with April’s peak in tax revenue driven by annual tax collections. The data indicates that subtle declines in cruise ship count appear to indicate a higher potential for a decline in general fund gross revenues four months later. In fact, the correlation is 54% and the R2 = 0.2908 between the four-month lagged ship count and actual monthly tax collections from January 2014 to June 2017.
This preliminary analysis of shipping tracking data suggests one of potentially many applications for using the data to assess municipal credit. The full dataset includes almost every category of large ship and could potentially provide new factors for economic and commodity price models. With some refinement and an expansion of the dataset, the ship data could potentially provide an advanced indicator that can be added to existing economic models to determine when Puerto Rico’s dire situation begins to change course.
Market makers are an integral part of a functioning municipal bond market and the quotes they send out every day to their clients provide a gauge for several aspects of liquidity. A very simple analogy would to be look at the quote data as a dealer’s mass marketing initiative and the trades as their actual revenue. A dealer needs to be careful about the quantity and quality of their broadly distributed quotes, as trading partners don’t like to see bids too far below the market or offers well above the market.
We reviewed daily bond dealer quote depth for the first half of 2017 for the approximately 201,000 unique bonds quoted and determined that 74% of the daily quotes reviewed had only one dealer quoting the bond on a given day. However, the remaining bonds had a much higher likelihood of trading with size on a given day as the number of dealers quoting a bond increased. Data indicates that there is a direct relationship between the number of dealers quoting a bond and the probability of an institution trading the bond in size (≥$500K trade size) that same day (Figure 3). The data indicates that the likelihood of trading doubles from 3% for single quoted bonds to 6% when a bond is quoted by two dealers. The trade percentage increases to 24% in the case of bonds quoted by five or more dealers, but we must note that only 0.8% of the quote depth dataset falls into that cohort.
Figure 4 compares the 2017 year-to-date daily count of bonds being quoted by five or more dealers to 10-year treasury bond yields. The number of bonds quoted by five or more dealers daily is admittedly low (typically less than 300 per day) compared to the vast size of the municipal bond market, but nevertheless it does have potential for use as a liquidity and herding metric. This year’s data indicates declines in liquidity near holidays (as expected) and above-average dealer quote depth coinciding with some of the highest and lowest treasury bond yields of the year.
There is currently no timely source of U.S. regional migration data that includes location and income strata to assist municipal bond investors in making an informed investment decision. The IRS publishes one of the best datasets focusing on population and income migrations; however, the data is typically released on a two year delay, so it fails to identify very recent demographic shifts that could prove to be very valuable from an analyst’s perspective.
One potential proxy for population migration is the changes in U.S. auto vehicle registrations, given that as of April 1 there were just over 269 million cars and light trucks registered in the U.S. according to the automotive team at IHS Markit. The group maintains an extensive history of auto ownership trends, which includes the ability to determine not only how many cars, trucks, or motorcycles left a state, but can also show the destination state, the model and model year of the vehicle.
The migration of new luxury vehicles between states is one potential gauge for the movements of higher income and net worth individuals among states. The movements of this group of vehicle owners could be a proxy for the number of senior managers and executives following companies to states with fast growing industries or owners moving or expanding to more business-friendly states, as an example.
Figure 5 shows 2016 net migrations of new (2014 and newer) luxury vehicles per 100,000 of state population. The data indicates that New Jersey had the worst net migration with 100 new luxury autos leaving the state per 100,000 residents and Massachusetts the best at 83 autos entering the state per 100,000 residents in the timeframe. It is worth noting that the data indicates that the entire New England region reported positive migration on the higher end of the auto market, while Florida had a modest negative migration. We also examined the ratio of new luxury vehicles entering versus leaving a state (Figure 6) and Idaho, Massachusetts, Texas, and Washington reported the most favorable ratio, with 2.3 new luxury cars entering those states for every one that leaves, while Ohio and Michigan were tied at the lowest ratio of 0.4 and California not too far behind at 0.5.
We note that vehicles that moved into a state as a rental fleet vehicle were not included in this analysis, but the data does include some autos that were fleet vehicles in the original state and then transitioned into personal vehicles in the destination state, based on ownership records. In particular, Oklahoma and any vehicles entering from the state were excluded from both charts, because of its unusually high proportion of rental fleet cars as a result of a large rental car company headquartered in the state. In addition, there will also be some degree of autos moving between states based on cross-border sales situations where the owner did not move with the vehicle.
Future analyses would include linking estimates of average owner income to model and year to further refine population migration estimates to determine which state economies are growing or contracting the fastest, and to determine to which states their residents are relocating.
The results of the cruise ship, dealer quotes, and luxury auto analyses provide only a starting point in their current forms. The future lies in money managers’ ability to effectively combine their expert knowledge of the municipal bond market with alternative, economic, and financial data to formulate better investment decisions. The quest for new alternative data with true value will always be a moving target, as even if there was one set of alternative data that provides all the answers today, the factor’s value would rapidly deteriorate once others include it in their quantitative models. The most sophisticated municipal bond investors of the future will be designing and constantly refitting models that combine hundreds or even thousands of factors in determining the correct price for risk and improve their ability minimize losses from events like Detroit’s and Puerto Rico’s defaults.