Friday, May 28, 2010

Fidelity Customers and iShares ETF's: the Impact of Commission-Free Trades

On February 2, 2010, Fidelity announced that their clients would be able to buy and sell a select group of 25 iShares ETF's commission-free, in addition to Fidelity's NASDAQ Composite Index Tracking Stock. The move follows a decision by Schwab offer commission-free trading on a select group of proprietary ETF's. Fidelity offers a larger number of commission-free ETF's than Schwab, plus the strong reputation of iShares.

In an earlier analysis of Schwab's commission-free arrangement, we studied the quality of trade execution at UBS, a major venue (technically, market center) for executing orders on Schwab ETF's. Even though UBS executes a large volume of orders for Schwab ETF's, other "ETF complexes", such as State Street, Vanguard, and PowerShares, exhibited better trade execution than the Schwab complex. This analysis of the iShares trade execution complex will follow a similar theme.

Morningstar appropriately asks in their commentary who really benefits more from the Fidelity-iShares arrangement: Fidelity or iShares? Their assessment concludes both parties win. Fundometry expands the audience to include investors who enter those commission-free orders. By looking at iShares ETF's which trade commission-free and those which do not (most of the iShares ETF's), we can compare execution quality with respect to the commission as well as time (before and after February 2010).

Fidelity is not a leading venue for trading iShares, or some of the other major ETF complexes. The following two graphs show how much volume was executed at NFS (Fidelity) versus other market centers. ETF complexes are selected from the top five traded ETF's, in terms of share volume over six months, at each market center. (Two graphs are better than one because the market centers range widely in terms of volume, hence the horizontal scales are different to accommodate the higher and lower volumes.) Someone entering an order through Fidelity to buy or sell an ETF (not only those from iShares) may hope the order is routed to another market center which handles a larger order flow, hence sourcing more liquidity and hopefully achieving better price execution. For those iShares which trade commission-free at Fidelity, the graphs show the volume for those respective tickers in a lighter shade, even though the actual trades may have incurred a commission if entered from brokerages other than Fidelity.



Interestingly, if an ETF order is entered from Fidelilty and gets routed to NFS, there appears to be a near certainty strong probability that the order will be executed at NFS. The following two graphs compare how much of the order flow was executed at the specified market center as opposed to being routed to another market center where a better price existed. NFS is not the only brokerage with a tendency to keep orders internal; Knight, Citadel, and Barclays also show no red bars.



Although NFS handles a relatively small volume of orders for iShares ETF's, order flow does not appear to route to other market centers which may offer deeper liquidity. Why? Perhaps because NFS does not need to route an order to another venue if it can execute at a price better than or equal to the National Best Bid and Offer (NBBO). Unfortunately, the following graphs, utilizing the same source data as the above graphs, do not readily support that explanation.



The above two graphs show the breakdown of share volumes according to execution quality during each month (October 2009 through March 2010): improvement (better than the quote), at the quote, and outside the quote (degradation). NFS (Fidelity), Knight, Citadel, and Barclays tend to execute orders themselves rather than route to other market centers (as stated earlier). Over the sampled period, Knight and Citadel exhibited noticeable price improvement for the tickers which Fidelity customers can trade commission-free. Barclays achieved price improvement to a lesser extent but executed a decent amount at the quote. In significant contrast, NFS showed a worsening trend in terms of execution quality. NFS substantially exceeded its competitors in executing outside the quote, a trend which dramatically increased in January and persisted through March. Why would Fidelity offer to not charge commissions for iShares ETF's for which it could not achieve better execution (even when charging commissions)?

Perhaps some more insight can be derived from looking at the other iShares ETF's which Fidelity still charges its customers to trade.



Once again, Knight and Citadel demonstrated a degree of price improvement, outpacing Barclays and NFS. While Fidelity charged a commission for trading the ETF's sampled in the above two graphs, the extent of price degradation was even higher than those ETF's which trade commission-free. In March, trades executed at NFS exhibited a higher rate of price improvement than in preceding months, and trades executed outside the quote subsided yet remained at a high level. Given that trading activity at these market centers are compared for the same group of ETF's and over the same period, the above graphs do not explain how Fidelity's customers (trading iShares ETF's commission-free or not) are winning with respect to execution quality.

Corresponding comparisons for some other major ETF complexes paint a similar picture: Direxion, PowerShares, and ProShares.

While many trades executed outside the quote, to what extent was price impacted? In other words, how far outside the quote did NFS fill orders? Since price impact (in dollars per share) is disclosed separately for trades where price improved and degraded, a weighted average, based on their respective share volumes, can be computed. The following graphs display the share volume in each market center for iShares ETF's which trade commission-free at Fidelity and the corresponding weight-average price impact ("WAPI").



In terms of WAPI, NFS once again stands out from its peers. NFS consistently achieved adverse (negative) overall price impact throughout the six-month period. As indicated above, the trend worsened after December. A peer with comparable volume, Barclays, also exhibited adverse WAPI but to a lesser extent than NFS and without significant trend. Among the largest market centers, BATS and Direct Edge exhibited adverse WAPI but to a much lesser extent than NFS. (Differences in the volume of specific ETF's at each market center explain some of these variations, but in general NFS did exhibit consistently more adverse WAPI versus BATS and Direct Edge among specific ETF's. If interested in more details, please submit a comment below requesting more information.)

The following graphs display the share volume in each market center for iShares ETF's which Fidelity charges its customers to trade and the corresponding weight-average price impact ("WAPI").



The performance across market centers is consistent with the iShares ETF's which trade commission-free at Fidelity. BATS, Direct Edge, Barclays, Knight Equity Markets, and NFS exhibit adverse WAPI. However, BATS, Direct Edge, Barclays, and Knight followed an improving trend over time. In terms of WAPI, only NFS achieved more adverse WAPI from late 2009 to early 2010.

Similar graphs for other ETF complexes similarly show NFS achieving increasingly adverse price impact over time, in contract to most peers. If the above graphs are not sufficient, click on the following ETF complexes to open a few more: Direxion, PowerShares, and ProShares.

In conclusion, while the arrangement between Fidelity and iShares does make tremendous sense from a marketing perspective, the quality of trade execution leaves much more to be desired. For financial advisors or long-term investors who prefer using ETF's, the price impact may not be meaningful enough to diminish the cost savings from zero commissions, as long as trades are not too frequent. Conversely, for day traders, the economics of avoiding a commission may not offset adverse price impact while executing an order at NFS (assuming that a Fidelity customer does not have the option of specifying at which market center to execute). As with any analysis of vast amounts of data, the results have varying relevance depending on individual investor circumstances. Nevertheless, one cannot help but wonder if NFS will improve its trade execution quality for ETF's - regardless of whether a commission is charged or not.

A few important details concerning the data utilized for this analysis:
1. The above graphs compare data gathered from Rule 605 disclosures from approximately a dozen market centers (trading venues). Those venues are selected based on the Rule 606 disclosure from National Financial Services ("NFS", an affiliated broker-dealer of Fidelity) which lists to which market centers Fidelity routed equity orders during the first quarter of 2010.
2. The Rule 605 disclosures for each market center (except for BIDS Trading) are used to analyze trade execution quality over the six month period from October 1, 2009 through March 31, 2010. Data at the individual ETF (ticker) level is aggregated according to the sponsoring firm (e.g. iShares, State Street, PowerShares).
3. The above graphs reflect only market and marketable limit orders. While other types of limit orders are frequently used, Rule 605 disclosures report execution quality for only market and marketable limit orders. To see a table summarizing activity by order type, click here.
4. Aggregated figures may not be ideally comparable across market centers and ETF complexes (sponsoring firms). Each market center has a different share of the trading volume for each ETF which impacts the comparison of execution quality across different market centers and during different months. Furthermore, not every market center has necessarily executed trades in every ticker within an ETF complex.
5. Weighted average price improvement may reflect some biases due to the differing prices of ETF's. Higher priced ETF's may have orders filled with greater deviations from the quote than a lower priced ETF. Conversely, lower priced ETF's may trade in larger share volumes than higher priced ETF's. The weighted average considers both the impact on price, in dollars, weighted by the corresponding share volume. Orders filled at the quote are included in the weighted average with an assumed price impact of zero.

Thursday, May 6, 2010

Treasury Temporary Guarantee Program: Update on Potential Costs to the Uninsured Investors


On September 18, 2009, the Treasury's Temporary Guarantee Program ("TGP") ended. This program provided a form of government insurance on participating money market funds. A few posts in November 2009 reviewed the manner in which payments to the Treasury (insurance "premiums") were calculated and charged by participating money market funds. The primary focus here is who paid for the insurance and who could have received any payments from the coverage.


While premiums were paid directly by the funds, the funds used their fees assessed to all money market fund investors, including any investors who did not qualify to receive a monetary benefit from the insurance. Investors in a participating fund would be covered by the TGP capped at the invested balance as of September 19, 2008 (the "Cut-Off Date"). In other words, even though premiums were charged at the fund level, coverage was defined at the specific account level and based on a specific time. This arrangement is akin to paying premiums for a mandatory employer-sponsored group insurance plan but not being covered if the employment started past a certain date.


Even people who invested in a money market fund as of the Cut-Off Date would not necessarily be fully covered by the TGP. If such an investor held $1,000 on the Cut-Off Date but subsequently invested another $2,000, only the initial $1,000 would be covered. In fact, coverage is not transferable across different funds. If the same investor redeemed the initial $1,000 which was covered and used the proceeds to invest in another money market fund, coverage would not apply to the new investment. An earlier post more clearly explains the details of how coverage was extended with respect to the Cut-Off Date.


The data to track who individually invested in money market funds, including when and by how much, is not publicly available. Even brokers possessing this information would probably require a non-trivial amount of effort and expense to retrieve such data. In order to demonstrate the monetary cost to investors who paid for the TGP but were not covered, we can utilize public data on cash flows for money market funds and derive some useful estimates from a model.


Based on the monthly sales and redemptions for registered money market funds, the model estimates how much of the fund balances originated from investments made before and after the Cut-Off Date. The results below exclude funds which invest primarily in Treasurys and Government debt. Several such funds decided not to participate in the TGP for the full term since, by investing in only Federal government-backed debt, the Treasury guarantee was implicit, hence additional insurance was deemed redundant. Therefore, the sample population of funds in this analysis comprises those which invest in CP, corporate bonds, municipal bonds, RMBS, ABS, and repurchase agreements, in addition to some possible amount of Treasurys and Federal agency debt. In order to expedite the modeling process, only funds with assets over $1 billion are included, and we assume such funds represent a majority of the population.


The results are broken down by three main scenarios. Across all scenarios, investors are assumed to redeem from money market funds in proportion to their balances. For example, if $100 million were invested in a money market fund prior to the Cut-Off Date, out of a total balance of $400 million for the same fund, then the model would assume that 25% of redemptions would be deducted from the $100 million portion. Of course, this "ratio" changes over the modeled time period, from September 19, 2008 through September 18, 2009, depending on how many investors purchase and sell a money market fund after the Cut-Off Date.


In the first scenario, the model assumes that purchases of a money market fund are as likely to come from investors who are covered (insured) by the TGP than those who are not covered (uninsured). Accordingly, in the numerical summary below, the split in Average Monthly Sales (82.6% insured vs 17.4% uninsured) closely matches the split in Average Monthly Balance (81.4% vs 18.6%). As highlighted by the green box, investor balances not covered by the TGP paid 9.6% of the total premiums in this model scenario.


In the subsequent graph, which shows the trend in the insured and uninsured balances over the modeled time period, the monetary cost to investors who are not covered by the TGP is proportional to the size of the red bars versus the green bars.


In the second scenario, the model assumes that purchases of a money market fund are twice as likely to come from investors who are covered by the TGP than those who are not covered. When comparing the split in Average Monthly Sales (45.4% insured vs 54.6% uninsured) against the split in Average Monthly Balance (65.7% insured vs 34.3% uninsured), we observe a greater tendency for sales/purchases to come from investors not covered by the TGP. As highlighted by the green box, investor balances not covered by the TGP paid 28.2% of the total premiums in this model scenario - equivalent to $188 million for the sampled population of $1.9 trillion in fund assets.


Accordingly, the graph corresponding to this second scenario shows a larger proportion of bars colored red than in the first scenario.


In the third and final scenario, the model assumes that purchases of a money market fund are four times as likely to come from investors who are covered by the TGP than those who are not covered. When comparing the split in Average Monthly Sales (25.7% insured vs 74.3% uninsured) against the split in Average Monthly Balance (49.9% insured vs 50.1% uninsured), we observe an even greater tendency for sales/purchases to come from investors not covered by the TGP. Investor balances not covered by the TGP paid a substantial 44.7% of the total premiums in this model scenario - equivalent to $298 million for the sampled population of $1.9 trillion in fund assets.


This scenario may be somewhat extreme but interesting nevertheless. In the corresponding graph, the proportion of red bars versus green bars shows the increasing amount of TGP premiums paid by investors who invested in a money market fund after Cut-Off Date.


While TGP premiums were paid to the Treasury only on specific dates, we assume that money market funds amortized that expense over time and spread the cost out beyond the date when premiums were paid. Regardless of when an investor made initial or subsequent investments, whether before/after Cut-Off Date and before/after premium payment dates, the fund assessed its fees consistently across all investors. Whether investors in a money market fund paid for some portion of TGP premiums after its termination on September 18, 2009 is a question beyond the scope of this analysis and beyond the scope of data available.


We cannot estimate an exact monetary cost to investors who, by definition of the TGP, would not have been eligible to receive payouts from the Treasury in the event a money market fund suffered losses from defaulted securities or otherwise could not fully repay investors. However, from these scenarios, we get a sense of the relative magnitude of this cost. The green box in each of the numerical summaries above indicates how insured and uninsured investors split the modeled cost of TGP insurance. Investors not covered under TGP might have paid $64 million under a reasonable scenario, to as much as almost $300 million in an extreme scenario. Without access to additional data from account-level records, one cannot perform a more accurate and conclusive analysis. (For a money market manager curious on how such an analysis would work, please contact the author at omer@ProteusFinancial.com)


This analysis should lead to further questions and potential ideas for how such a program could have been funded in a more equitable manner. Please feel free to post comments and feedback below. Check back for a future post showing how an actual money market fund (which invests in Treasury securities) participated in the TGP for the full term even though a vast majority of investors purchased the fund after the Cut-Off Date.

Friday, April 9, 2010

In Search of Price Improvement for Schwab ETF's


When Charles Schwab sold its capital markets affiliate to UBS in October 2004, the industry took notice of the impact: UBS gained significant market share in NASDAQ equities trading as well as the sizable equities order flow from Schwab's customers. Schwab does not receive rebates or other payments from UBS. Nevertheless, part of the consideration Schwab received for the sale was a commitment to route orders through UBS for eight years, starting in 2004.


Normally, Schwab charges a commission for trading equities and options. However, on November 3, 2009, Schwab announced the launch of their proprietary ETF's as well as their unprecedented $0 online commission rate for Schwab clients.


Does this no-commission/no-rebate arrangement have any impact on order flow or order execution? Comprehensive details are not easy to collect, but the following analysis is an attempt at providing some insights from data which is available.


This analysis utilizes Rule 605 disclosures for the following exchanges and ECN's ("market centers"): UBS, NASDAQ, NYSE Arca, BATS, DirectEdge, Instinet, Knight Capital, and Citadel. The sample period covers November 2009 through January 2010. Other market centers executed or routed orders for Schwab ETF's, but this sample accounts for 72% of the trading volume over the sampling period.




Below, we look at various statistics among some of these market centers with a specific focus on UBS.


(Note on the following graphs: Individual ETF's are aggregated according to their sponsoring firm into "ETF complexes". Although the selected ETF complexes have trading volumes larger and smaller than that of the Schwab complex, Schwab's ETF volume should fall roughly in the center of the distribution. Shown next to each ETF complex is the aggregate volume, over the sample period, expressed in millions of shares. The graphs do compare ETF complexes with varying amounts of liquidity, but one can isolate those with similar trading volumes by noting the figures in parenthesis.)


What happens when a broker (assumed to be Schwab) routes a Schwab ETF order through UBS?




We observe how many shares are traded at market centers other than UBS, out of all shares routed and traded through UBS. Few other ETF complexes get such exclusive treatment by UBS, certainly none as large as the Schwab complex. UBS traded fewer shares of First Trust, Rydex, and WisdomTree ETF's (10.4, 9.3, and 11.9 million shares respectively) than Schwab (17.3 million shares) yet routed more of their orders to other market centers.


What happens when a Schwab ETF order is routed through other major market centers?






If a market center cannot fill an order at the best price (national best bid and offer, or "NBBO"), then that center should route the order to another venue where a better price is available. On that basis, other market centers don't seem as capable as UBS in achieving best trade execution for Schwab ETF's. Perhaps this makes sense if the investors trading Schwab ETF's are mostly Schwab clients. As seen above, although NYSE Arca handles orders representing 2.1 million shares of Schwab ETF's, it routes almost 10% of those orders to other market centers where better prices are presumed to have been available.


How well did UBS fill those orders which it chose to handle internally (i.e. not route to other market centers)? Given the data available, we can look at how many shares were executed inside the quote (known as "price improvement"), at the quote, and outside the quote. The following graphs display the percentage of shares which fall into one of these "three tiers" of execution quality. Most orders are either market orders or marketable limit orders. Therefore, we report the two types separately since there may be inherent differences between the manner in which market orders and marketable limit orders are executed. One would especially hope for price improvement on market orders, but no worse than filling at the quote.


Given that UBS executed almost all Schwab ETF's orders internally, it is interesting to see how UBS fills market orders. Among complexes of comparable trading volume, Schwab ranked quite low on price improvement. UBS achieved the best price improvement for iShares ETF's, at about 75% of the total 228 million shares traded with market orders. First Trust ETF's, with a smaller volume of 9.1 million shares traded with market orders, still saw price improvement for 65% of that order flow. UBS achieved price improvement for Schwab ETF's for about half of the 14.7 million shares traded with market orders. The results for at the quote and outside the quote fills are respectable, but nothing special either.




For marketable limit orders, UBS handled Schwab ETF volume of 2.6 million shares. Price improvement ranked somewhat better across the same sample of ETF complexes, but not by much. Approximately 3% of shares traded with marketable limit orders saw price improvement, while almost all others were executed at the quote. UBS achieved the best rate of price improvement for Victoria Bay ETF's with a higher volume (24.6 million shares). ETF's from First Trust and Rydex, with 1.3 and 1.6 shares traded respectively, saw better rates of price improvement even though their volumes are smaller than that of Schwab ETF's.




The same statistics are available at other market centers. Results can be skewed depending on the order volume handled by each market center. Hence, the initial table, showing volume of executed orders by market center, may be a helpful reference.


In comparison to other ETF complexes, NYSE Arca executed market orders and marketable limit orders for Schwab ETF's with modest price improvement, which falls near the lower end of the range. Note that the volume of market orders is very small, only 0.1 million shares of Schwab ETF's executed and no more than a million shares for any other ETF complex. The volumes for marketable limit orders was higher, with 2.0 million shares of Schwab ETF's executed, and Schwab ranked roughly in the middle of the distribution.






Similarly, Schwab ETF's traded on NASDAQ with very little price improvement, but nevertheless ranked near the middle of the range among other ETF complexes with comparable volumes. NASDAQ did not disclose any statistics for market orders.




For orders executed at BATS, Schwab ETF's received the most-frequent price improvement for marketable limit orders in comparison to peers, while price improvement seldom occurred for market orders. Out of 0.8 million shares traded in Schwab ETF's with marketable limit orders, BATS achieved price improvement for approximately 12% of the order flow. While this percentage is much lower than the rate of price improvement achieved by UBS, Schwab ETF's ranked better than other comparable ETF complexes for marketable limit orders executed at BATS.






While digesting these graphs, consider these additional factors.


1. The trading activity spans only three months, although all market centers and ETF complexes are compared over the same period. Schwab launched its proprietary ETF's in November 2009, hence relevant activity does not exist for prior months.

2. Only online orders entered through Schwab are commission free. Orders entered through other channels (e.g. phone) are not advertised as commission free, even though those orders would likely be executed through UBS. The available data does not provide any information on the order entry method. Hence, this analysis assumes that the majority of Schwab ETF's orders routed through UBS were entered online.

3. Other attributes may differ between orders routed through UBS and orders routed through other market centers. No details are available on actual limit prices on limit orders, and the specific execution times and applicable NBBO's are not provided. However, further analysis can be performed on the amount of price improvement at the ticker level.

Execution quality is challenging to track and measure. The summarized data used in this analysis attempts to compare execution quality among major market centers at the level of ETF complexes. While not digging into finer levels of comparison (ticker level or specific month), this analysis demonstrates that Schwab ETF orders routed through UBS almost exclusively get executed at UBS and generally with less frequent price improvement than orders executed at UBS for other ETF complexes. Nevertheless, the other market centers in the sampled data exhibited even less price improvement.


UBS probably has an advantage versus other market centers due to its sizable portion of order flow for Schwab ETF's, even though other ETF complexes with smaller volumes (such as Claymore, WisdomTree, and Rydex) achieved more frequent price improvement at UBS. On the other hand, UBS achieved price improvement for other ETF complexes (such as iShares, Vanguard, PowerShares) to a greater extent than for Schwab ETF's even though a larger portion of orders were routed to other market centers.


Would paying a commission in order to trade an iShares ETF have resulted in more cost-effective execution than paying no commission to trade a Schwab ETF? The answer may depend primarily on order size and the frequency of trading. This analysis is not exhaustive and cannot be conclusive, but the sampled data demonstrates that paying no commission when trading an ETF might not always result in the lowest overall execution cost.