About a month ago, we spent a full week highlighting research conducted at our firm that shows the degree to which investors are harmed by excessive markups in municipal bond trading. In the paper, our colleagues argue that low-cost improvements in disclosure requirements could largely eliminate these transfers of wealth from taxpayers and investors to the brokerage industry.
After the research was completed, we began thinking about other ways investors gain exposure to municipal bonds. For example, investors can buy shares of a fund whose assets are municipal bonds. As a result, we began looking into municipal bond funds to determine the typical markup charged on their trades.
To begin our analysis, we looked at one of the largest municipal bond exchange traded fund (ETF): SPDR® Nuveen Barclays Short Term Municipal Bond ETF (SHM). We obtained the daily holdings for the first half of 2013 via Bloomberg and determined dates on which the holdings of the fund changed (implying a trade).
Using these lists of bonds held by the fund, we obtained the trading activity for each municipal bond from the Electronic Municipal Market Access (EMMA) website. At this point, we attempted to reconcile the trading activity implied by the fund holdings with the trading activity given by EMMA. For example, if the fund increased the position in a given bond, the EMMA data should reflect a "Customer bought" order for the same amount around the same time.
Let's look at one example that is consistent with our intuition. On February 14, 2013, SHM increased its holdings of a municipal bond with a CUSIP of 208418NN9 by $695,000 according to Bloomberg. EMMA has the following data for the bond around this date.
There is a transaction on February 12 in the same amount that is denoted as a "Customer bought", consistent with the interpretation that the fund bought the bonds on or about this date.
Now let's look at another time that SHM increased its position. On March 22, 2013,SHM increased its holdings of 403755L74 by $1,325,000. EMMA has the following data for the bond around this date.
The only trade occurring around this date with the correct trade amount is a "Customer sold" order on March 20, 2013. Given the previous example, it's not clear how this is consistent.
Finally, on March 12, 2013,SHMincreased its holdings of 613340V39 by $5,000,000. The following image is all of the trading activity EMMA has for the bond.
You can see that there is no "Customer bought" or "Customer sold" trades to match with this change in position -- in fact, there are no trades at all within several months of the transaction date!
This same phenomenon also exists for fund sales: sometimes it is reflected with a "Customer sale" in the EMMA data, sometimes it is reflected in a "Customer bought," and sometimes it is not reflected at all. We have also attempted to piece together a trade that would match the fund transaction from several smaller trades in the EMMA data around the same time, with limited success. It thus appears that the EMMA data and the fund position data from Bloomberg are sometimes inconsistent.
Earlier this week, the Municipal Securities Rulemaking Board (MSRB) announced that "it is seeking additional public input on transforming the public display of real-time trade prices on its Electronic Municipal Market Access (EMMA)". One question that our analysis raises is how complete the EMMA data really is, and if it fully reflects ETF transactions. This information is important for retail investors, who may want to know whether they are charged larger or smaller markups than those charged to institutional investors.
_______________________________________
UPDATE(12:15PM): The last bond we referenced (CUSIP: 613340V39) changed from the SPDR Nuveen Barclays Municipal Bond ETF (TFI) to the SPDR Nuveen Barclays Short Term Municipal Bond ETF (SHM), both held in the SPDR Trust. The 2007 5% Montgomory County, MD General Obligation bond can be found in TFI's portfolio in the December 31, 2012 report and in SHM's portfolio in the March 31, 2013 report . Therefore, the bond may have been transferred from one SPDR portfolio to another. Perhaps this kind of transaction is not reflected in the EMMA data, but we'll keep tracking this issue down.