www.indexuniverse.eu/joi trading the index September/October 2012 Trading Costs And Index Design Konrad Sippel Index Tradeability Xiaowei Kang and Daniel Ung An Index Impact Analysis Gareth Parker ETF Creation/Redemption Models Ruben Kersbergen and Christiaan Scholtes
Contributors Gareth Parker Gareth Parker is a senior director of index research, design and development for Russell Indexes EMEA, Russell Investments. Gareth works closely with index users to understand their index requirements and to develop Russell indexes that match their exact needs. In the past eighteen months, Gareth has been responsible for the design and development of a number of indexes. In 2011, he designed the BATS Chi-X Russell Europe Index Series (CHERI Series) which represents broad pan-european and eurozone exposure. In 2012 he worked closely with the trading community to design the Russell Europe SMID 300 Index, containing the 300 most rapidly tradable constituents from the SMID- cap developed Europe opportunity set, which NYSE Liffe has listed on its Bclear platform. Gareth is a regular speaker on the international European finance conference circuit and acts as a media spokesperson for Russell Indexes. He has particular interest in the indexation of non-linear payoffs, less correlated markets, and of complex strategies. 4 September/October 2012
Index Tradeability Rules An impact analysis By Gareth Parker 22 September/October 2012
FROM INDEX UNIVERSE TO TRADEABLE BASKET Over the years the way in which indices are used has changed. Whereas initially indices were simple average measures of market movements, they have now become full and accurate measures of the investable opportunity set, as well as tools on which portfolios and index basket trades are precisely modelled. In order to accommodate this change, rules have been incorporated into index methodologies to reduce the problems caused by constituents that cannot be traded efficiently. Historically these rules were focused more on the needs of investors, who were aiming to trade cheaply. More recently, index designers have also tried to address the needs of sell-side firms, who tend to want to trade quickly. The first major step towards ensuring the tradeability of indices arguably came with the widespread adoption of market capitalisation as a weighting methodology. The subsequent move by most index providers to weight constituents by free float was also motivated primarily by concerns over liquidity. Whilst index providers free float methodologies tend to differ in minor ways, each is attempting to ensure tradeability by removing from the indices that percentage of each constituent that is not available for investment. Shares considered as not part of a company s free float are typically in the hands of company founders or other controlling shareholders, or not available for trading as a result of other restrictions. In the following analysis, we examine a variety of index weighting/screening rules from the perspective of those rules impact on tradeability. This is done with a view to identifying the shortcomings and advantages of each approach, given the different objectives of index users. In particular, we explore the following methods: 1. Price-Weighting. The original index weighting methodology, as used in Charles Dow s first indices. 2. Full Capitalisation-Weighting. Each constituent is weighted according to its full, unadjusted market capitalisation. 3. Trading Day Screening. A simple screen that requires trading in each constituent to have occurred on a given number of trading days within a sample period. 4. Free Float-Weighting. The percentage of each individual constituent s shareholdings that are considered not to be freely available for trading is removed. 5. Free Float-Screening. The removal from the index of those constituents with the smallest free floats. 6. Traded Value Screening. Removing companies with the lowest average absolute trading value. 7. Speed To Trade/Velocity Screening. The company s free float capitalisation is compared with its average traded value. Constituents that would cause a notable increase in the time to trade an index basket are excluded. This approach is a relatively recent index development, seen particularly in indices developed for use by sell-side firms. 1 8. Equal-Weighting. This methodology is not specifically focused on tradeability, but its recent popularity makes it worth including in the analysis. MEASURING INDEX TRADEABILITY Our analysis focuses on the impact of different index methodologies on three of the measures most favoured by market participants to measure the tradeability of a basket of stocks: weighted average dollar-denominated value of trading (ADDTV), market capacity and speed to trade, as defined below. Taken in isolation, none of these measures represents a sufficient, stand-alone gauge of index liquidity. For example, there is no absolute capacity value or speed to trade that is unacceptable, except in individual cases where the index designer s client has specified such a requirement. However, the three measures we have chosen for this analysis provide a useful overall picture of an index s tradeability. ADDTV is presented as a partial proxy for the cost of trading, as there is a direct relationship between ADDTV and spreads/trading costs; Market capacity is defined as the largest index basket trade that can be made without breaching 20% of any individual constituent s ADDTV; Speed to trade analyses the time taken to trade each individual constituent by comparing its ADDTV with the actual trade size required in that constituent as a part of a given index basket trade. The overall speed to trade figure of an index is determined by the largest speed to trade figure of any individual constituent, meaning that the slowest to trade constituent determines that overall figure. In the analysis below, we assume a US$100 million notional trade size in each case. COMPARING LIKE WITH LIKE Rules looking to increase the tradeability of an equity index tend to be simply screens that require each constituent to meet either an absolute or a relative measure of tradeability. As these levels of tradeability are set by the index designer in order to achieve particular results, the minimum trade days requirement of one index is, for example, likely to be different from that of another index. In any case, two different indices may be using different measures entirely, making it difficult to ensure comparability. In order to attempt a fair, like-for-like comparison, we have in each case first used the tradeability rule being investigated to remove the lowest-ranked 2% of the universe of securities. This figure represented the largest percentage that could be removed in all cases, since 98% of the constituents in the selected universe traded on every day during the period under review. 2 www.indexuniverse.eu/joi September/October 2012 23
Then, to create as realistic a picture as possible, after incorporating the tradeability rule we also removed the bottom 1% by capitalisation of each index s least traded constituents. This was done to account for the ability of institutions, whether investors or market-making firms, either to use portfolio optimisation techniques (and thereby avoid trading the index s least liquid constituents), or to trade on dark pools (where fund managers have anecdotally reported being able to trade more cheaply than is implied by ADDTV figures from constituents primary exchanges). To determine which index constituents to remove, we measured each constituent s ADDTV, which is defined as the total value of trading in each constituent over one year, weighted by each constituent s index weight. 3 For each tradeability rule under review, these actions removed between four and six of the index constituents. In the analysis whose results are shown in Figure 1 we assume that, after the adjustments described above, the resulting portfolio was in each case replicated in its entirety according to the given methodology. ANALYSIS For the purposes of this analysis we used developed European companies, ranked from 301 st to 600 th by full market capitalisation, using data from 31 May 2012. This universe of stocks has notable liquidity problems, as indicated both by the data presented here and by the existence of indices such as the Russell Europe SMID 300, Russell Eurozone SMID 150, STOXX Mid 200 and STOXX Small 200, each built with the objective of improving tradeability. With the exception of the equal and price-weighted methodologies, each of the weighting or screening options has been applied to this constituent set. The final constituents are then weighted by market capitalisation to allow a comparison of the individual efficacy of each index methodology. The results are shown in Figure 1. COMMENTARY ON RESULTS Price-Weighting First, we found that, unsurprisingly for an index that sets no tradeability criteria, a portfolio following the rules of a basic price-weighted index had a very poor investment capacity of US$11 million. This was despite the removal of the bottom 1% of constituents by ADDTV, which, as noted above, was done to emulate market participants ability to optimise for some tradeability issues. The weighted average trading value of the constituents for the basic price-weighted index was US$13.3 million, the lowest of the index methodologies surveyed in Figure 1. It would also take 45 days to trade a US$100 million index basket. In addition, the methodology caused some significant country skews from a starting point of a free float market capitalisation methodology: the UK s index weighting, for example, dropped from 33% to 27%. Full Capitalisation-Weighting Weighting constituents by their full market capitalisation was pioneered by S&P for its 100 index of the largest US stocks. Counter-intuitively, using this approach in our analysis did not significantly increase index tradeability. In Figure 1 the weighted average ADDTV of the full capitalisation-weighted approach increased only to US$14.8million from the US$13.3million of the priceweighted index; the index s capacity also rose only modestly, to US$13.8 million, and the days to trade figure improved only to 36 days from the full capitalisationweighted index s 45 days. This outcome is due to a rapid worsening of trading speed in the smallest ten constituents, particularly the last three, each of which required over ten days to complete a trade of the required size. Trading Day Screening Unsurprisingly, by removing most of the companies that are so illiquid that on some days they do not trade at all, this screen dramatically improves the overall length of time taken to conduct a basket trade in the index s constituents, for example when compared with the simple market capitalisation-weighted approach. The weighted average trading value of the constituents is also improved. Most notably, the investment capacity in this option increased to US$108.5 million. Although this figure is higher than the capacity of the previous two index methodologies reviewed, it is still arguably insufficient in a market (European smalland mid-cap stocks) where basket trades of U$100 million or more are not uncommon. Free Float-Weighting The free float-weighting of constituents provides a stepchange in the tradeability of the index. In Figure 1 the index s capacity increases to US$1.7 billion and its weighted average traded value increases further, again to US$17.3 million, with the time taken to complete a US$100 million trade dropping to 0.3 days. The free float weighting methodology achieved this improvement while removing only one more constituent than the trading day screening. Free Float-Screening Some index providers now not only weight constituents by free float, but also screen for a minimum required free float percentage. In our analysis we have used a minimum free float requirement of 25%, similar to most providers existing rules. Unsurprisingly, the complete removal of companies with low free floats, which by definition are likely to be relatively illiquid, nearly doubled the investment capacity of the resulting index portfolio from the basic free floatweighting methodology, from US$1.7 billion to US$3.3 billion. The time to trade was also reduced further, to 0.15 days. This improvement, however, came at the cost of removing considerably more constituents: 39 in all from the starting universe of 300 stocks. This number is higher than for any of the other options, skewing the country weightings noticeably. 24 September/October 2012
Figure 1 Effect Of Liquidity Weighting And Selection Methodologies On Tradeability Price Wt Full Cap Wt Trading Days Screen FF Cap Wt FF Cap Screen Traded Value Screen Velocity Screen Equal Wt Capacity@20% (US$) 10,897,536 13,804,264 108,461,973 1,680,538,022 3,246,068,803 3,285,238,864 4,205,404,070 11,044,800 Overall days to trade 45.88 36.22 4.61 0.30 0.15 0.15 0.12 45.27 Weighted avg. DTT 0.38 0.31 0.09 0.04 0.04 0.04 0.04 0.37 Minimum ADDTV (US$) 7,363 7,363 95,600 547,161 1,325,184 1,421,054 95,600 7,363 Weighted avg. ADDTV (US$) 13,513,150 14,826,914 15,464,320 17,268,046 17,733,927 17,612,651 17,618,059 13,332,974 Avg. trading days 99.6% 94.8% 100.0% 100.0% 100.0% 100.0% 100.0% 99.6% Weighted avg. trading days 99.6% 98.6% 100.0% 99.0% 100.0% 100.0% 100.0% 98.3% Final nr. of constituents 296 296 284 283 261 275 281 296 Austria 4.1% 4.2% 4.2% 3.0% 2.5% 2.8% 2.8% 4.0% Belgium 2.7% 2.7% 2.9% 2.6% 2.7% 2.5% 2.5% 2.7% Denmark 3.7% 3.8% 3.4% 3.1% 3.1% 3.0% 3.0% 3.7% Finland 2.7% 2.9% 3.0% 3.6% 3.7% 3.7% 3.7% 2.7% France 11.8% 13.1% 13.5% 10.9% 10.8% 11.0% 11.0% 12.0% Germany 9.8% 11.0% 9.4% 8.5% 8.3% 8.4% 8.5% 10.7% Ireland 2.7% 2.4% 1.4% 2.9% 3.0% 2.8% 2.9% 2.7% Italy 4.4% 4.0% 4.2% 3.0% 2.9% 3.1% 3.1% 4.3% Luxembourg 1.0% 0.8% 0.9% 0.6% 0.6% 0.6% 0.6% 1.0% Netherlands 4.1% 4.3% 4.5% 4.7% 4.9% 4.8% 4.8% 4.0% Norway 2.7% 2.4% 2.5% 2.1% 2.0% 2.0% 2.0% 2.7% Portugal 1.0% 1.0% 1.1% 0.6% 0.5% 0.6% 0.6% 1.0% Spain 5.7% 5.8% 6.1% 5.2% 5.1% 5.3% 5.3% 5.7% Sweden 5.4% 5.0% 5.2% 5.6% 5.6% 5.6% 5.6% 5.3% Switzerland 10.5% 10.4% 10.7% 10.5% 10.3% 10.2% 10.1% 10.3% UK 27.7% 26.2% 27.1% 33.3% 34.1% 33.8% 33.5% 27.3% Source: Russell Indexes, Bloomberg Traded Value Screening Screening for the traded value of constituents, rather than for low free float, gave a similar result. Where traded value screens are used they tend to be set in terms of an absolute minimum figure in nominal terms. This approach introduces the likelihood of index turnover being driven by changes in market conditions and market valuations, rather than by changes in constituents relative positions on a tradeability measure alone. Screening for a minimum free float percentage avoids this turnover. We therefore consider that the free float methodology makes more sense than traded value screening as an index methodology. Speed To Trade/Velocity Screening Screening by speed to trade is a measure that has been introduced to isolate and exclude index constituents that cause a significant increase in market impact and exposure to market volatility. Unsurprisingly, in our test this approach delivered by far the highest available investment capacity, without a significant impact on weighted average traded value. This approach also resulted in the quickest speed to trade of the methodologies surveyed, at 0.12 days. As this option does not screen for company size, unlike the more standard approaches of free float and traded value screening, one might expect country weightings to be skewed significantly. However, in our analysis country weightings did not appear to deviate much from the capitalisation-weighted starting point. This was due in part to a smaller number of constituents being removed under this approach, when compared with the other screening options. Continues on page 45 www.indexuniverse.eu/joi September/October 2012 25
Continued from page 25 It is apparent, though, that the results experienced when using this approach will depend significantly on the screening level used. Velocity screening is more prone to removing constituents from across the market capitalisation range, unlike size-focused screening methodologies, which tend to remove smaller index constituents as a result of a focus on capitalisation or traded value. 4 Equal-Weighting An equal-weighted index approach has clear limitations in less liquid markets. The overall capacity and time to trade for this option turned out to be very similar to that of the price-weighted option in our test, despite the removal of the lowest 1% of the index capitalisation as measured by ADDTV. The substantial changes to country weights could of course be considered a positive, given that equalweight indices are marketed as one way of correcting the perceived biases in capitalisation-based portfolios. However, the limited investment capacity shown for this approach suggests that an equally weighted index should only be considered as the basis for an investment portfolio or for small investment amounts. In other words, we find that this approach suffers from potentially poor tradeability. CONCLUSION It is evident that weighting by free float-adjusted market capitalisation is a simple and very effective method to provide indices that are suitable for investment. Beyond that, however, screening either by minimum free float, traded value or both may provide additional tradeability gains, and indeed many indices use one or more of these screens in their construction methodologies. Our analysis suggests that speed to trade (velocity screening) offers notable further advantages by focusing on the requirement to complete large trades rapidly, thereby reducing market risks. The various potential country, sector and other skews that result from a deliberate focus on tradeability represent a trade-off, however. When the index provider removes or reduces the weight of an index s less tradeable constituents, the make-up of the new index starts to differ significantly from the original selection universe and the new index s performance also becomes less correlated with that of the starting set. Historically, broad benchmark indices were standard and were designed to measure the overall investment opportunity set accurately and fully. As such indices typically set no or very limited tradeability requirements, they contained the vast majority of the potentially eligible companies. By contrast, many more recent indices set more stringent criteria for tradeability. By removing more stocks from the eligible universe, they automatically incur increased country, sector and other skews. Although these skews are an accepted trade-off, index designers will generally endeavour to reduce them as much as possible. In our analysis we have attempted to provide an overview of how index methodologies may be designed to improve tradeability. We show that some newer methodologies for promoting index tradeability come at a cost of increased deviations from the original benchmark. As a result, such index approaches may be likely to be used primarily in indices designed for use as the basis of derivative products, where tradeability takes precedence over correlations with the full starting set of securities. Most importantly, it is evident that, with a broader set of design options available to the index provider, indices can be designed to suit any end-user s requirements; that is, whether the index user is focused on the full representation of the investment universe, or on the maximum tradeability of the index s constituents. 5 Endnotes [1] For an example of an index built using this methodology, see the Russell Europe SMID 300 Index Construction & Methodology document, available at http://www.russell.com/indexes/documents/russell-europe-smid-300-index-construction-methodology.pdf [2] The analysis was checked by repeating the exercise but removing stocks that are more than one standard deviation away from the mean value of the tradeability criterion in question. The results were very similar. [3] For more information on the definition of ADDTV please see the Russell Global Indexes Construction & Methodology document, available at http://www.russell.com/indexes/documents/global_indexes_methodology.pdf [4] For a close examination of an index built using this option, see: An Introduction to the Russell Europe SMID 300 Index paper by Parker 2012, available at: http://www.russell.com/indexes/documents/research/introduction-to-russell-europe-smid-300-index.pdf [5] An interesting further discussion can be found in Christopherson 2012: The Making of a Better Benchmark, available at: http://www.russell.com/indexes/documents/research/making-of-a-better-benchmark.pdf For more information on the Russell Europe SMID 300 Index and other products, contact: Tel: +44 (0) 20 7024 6600 Email: index@russell.com www.russell.com/indexes www.indexuniverse.eu/joi September/October 2012 45