Sixth Annual State Street
Global Markets Research Retreat

1 $1 Trillion in Portfolio Transitions

2 Investor Styles and Portfolio Flows

4 Fixed Income Portfolio Reallocation Analysis

6 State Street Investor Confidence Index

7 Sovereign Bond Flow Indicator (SBFI)

State Street Associates presented new research into investor behavior and portfolio and risk management at this year’s conference, which included the participation of pension funds, hedge funds, asset managers and academics from all over the world.

Features invited speakers at this year’s event were Robert Merton, Professor at the Harvard Business School; Martin Feldstein, Professor of Economics at Harvard University; Robert J. Shiller, Professor of Economics, Yale University; Professors Randy Cohen and Andre Perold of the Harvard Business School and Steven Levitt, Professor of Economics, University of Chicago and Author of “Freakonomics”.

Lessons Learned from $1 trillion in Portfolio Transitions

Mark Kritzman and Sebastien Page

Sebastien Page
State Street Global Market’s industry-leading Transition Management business has generated a unique and rich database of some $1 trillion in self-financing portfolio reallocations, comprising 2,992 transition events and 2.4 million transactions. These reallocations resulted from specific decisions by pension plans to change managers or asset mixes.

This study, which is being undertaken with Simon Myrgren of State Street Associates, and which is still underway, will be the first-ever public domain examination of self-financing implementation shortfall. The simplest way to define implementation shortfall is simply to calculate the value of a target portfolio at the moment when the decision to reallocate was taken and then, after the reallocation event, to calculate the actual cost, reflective of opportunity cost and market impact.

What is the real cost of trading? Andre Perold, who is also speaking at this event, suggests that it is the difference between performance on paper and in reality. The target portfolio, in effect, becomes the benchmark. Anything that happens on the way to that benchmark - commissions, taxes, bid ask spreads, opportunity costs and market impact are calculated as implementation shortfall.

Briefly stated, opportunity costs represent prices moving away from the investor as they implement the portfolio-prices moving up as securities are bought, or down as they are sold. Market impact often arises because a transition may constitute a large proportion of a security’s daily volume - the transition may be moving the market. While there have been many studies of market impact, there have been none on the uniquely difficult implementation shortfall that arises from a transition management event - when one portfolio is liquidated and another acquired in the same cluster of transactions.

In calculating implementation shortfall we sum all the costs for the securities we are selling and buying and simply divide by our initial portfolio value. This gives us implementation shortfall at the portfolio level. But in this example we are talking about only one portfolio reallocation. For transition managers who are managing multiple portfolio reallocations, issues like correlations, risk and tracking error makes life both complex and interesting.

Trading costs vary according to method. In-kind transfers happen when the same assets are in both the legacy and target portfolios. Internal crossing - our most cost-efficient method - can arise when the transition manager has multiple clients that are trading with each other or when the transition manager is trading on behalf of multiple clients. For example at State Street, with transition management clients and large index funds constantly rebalancing their portfolios we can internally cross some 36 per cent of volume. Plan A is selling a security, Fund B is buying, and the trade can take place off-market with no commission or market impact. External crossing is between buy-side and buy-side, for example over electronic networks. Least desirable of course is open market trading - a sort of trade venue of last result for a large transition.

Not surprisingly, implicit costs explain most implementation shortfall. In analyzing sectional correlations in some 400 transitions, we discovered that opportunity costs and market impact accounted for some 97% of total costs. And when tracking error increases, so does the risk of shortfall. The question then arises: can we come up with a measure similar to a Sharpe ratio (a risk-adjusted measure of return calculated by dividing the annualized excess return of a portfolio by its total risk-developed by William P. Sharpe) with which to judge the efficiency of a portfolio reallocation?

$1 Trillion in
Portfolio Reallocations,
2,992 events and
2.4 Million Transactions

We are now at the research stage of developing just such a measure. The concept would be to define risk adjusted shortfall as the realized explicit cost divided by the expected explicit costs and adjust the rest for a market impact function and tracking error risk-the transition manager’s Sharpe ratio or information ratio, if you will. Now that we have highlighted the importance of liquidity and tracking error in trading, the question is, how can transition managers improve their risk-adjusted shortfall score? The answer, not surprisingly, is through optimized trading.

Mark Kritzman
Reducing opportunity cost is the primary challenge in a portfolio transition. These costs should be zero across many transactions. But for any single transaction this is the greatest single risk - that the securities in the legacy portfolio will fall in price before you have an opportunity to sell them or that the securities in the target portfolio will go up before you can acquire them.

There are a variety of algorithms in the literature for reducing opportunity costs. But they don’t address the unique features of the challenges faced in transition management because in the main, they just deal with one-sided trades. The principle challenge before transition managers is to reduce net volatility between the legacy and target portfolios. Traditional trading algorithms don’t address a central feature of portfolio transitions - that they must be self-financing; for every dollar’s worth of securities sold, a dollar’s worth must be bought.

Working with ideas that have grown out of portfolio optimization algorithms developed by William Sharpe 20 years ago, we came up with an algorithm that does the following

• Defines tracking error as a function of the covariance matrix of the existing securities in the legacy and target portfolios and potential trades.

• Calculates the partial derivative of tracking error with respect to proposed trades to determine optimal sequence of trades.

• Sets derivatives of securities to be sold equal to each other and securities to be bought equal to each other to determine optimal trade size.

• Chooses the smaller trade to avoid trading a suboptimal amount in the other security.

• And then repeats this process until the transition from the legacy to the target portfolio is complete.

As far as managing market impact goes, there is in the literature a great deal to suggest that we can directly trade off market impact and opportunity costs. But in the case of transition managers this is impossible because of the self-financing constraint. So what we do instead is consider market impact as a way of telling us how to partition these trades to optimally reduce opportunity costs. And then we use the standard market impact management techniques which require us to make some assumptions about the number of shares, average daily volume, price, market value, etc. When we apply this market impact function to our algorithm, it simply shows that we partitioned the trades into smaller units.

The greatest risks to portfolio transitions generally are opportunity cost and market impact. For an individual transition, it’s probably opportunity cost. We’ve analyzed these costs thoroughly, and we’ve come up with a method for risk adjustment that scales implementation shortfall as a function of illiquidity and tracking error. In order to control these risks, we have introduced an algorithm that minimizes tracking error between the legacy and target portfolios as a function of the sequence and size of trades, taking market impact explicitly into account. Our algorithm reduces opportunity cost by 40% compared to the industry norm.

Investor Styles and Portfolio flows

Kenneth Froot
Whenever we consider investment flows an obvious question arises - what is the value of investment flow information, when for every buyer there is a seller? For several years we have been undertaking research around net flows emanating from institutional investors representing some $10 trillion in investments. And when you consider that we’ve been looking at this information for several years, our sample quickly aggregates to over $50 trillion worth of institutional investment. And these investment flows really represent yet more trillions because we see only those investment funds administered by State Street. Presumably investment decisions reflected in our information are also reflected in funds administered at other institutions.

What we are trying to understand is how investors and their strategies interact in the marketplace. What is the importance of this interaction? What is the capacity that they bring to their risk? And what is the urgency of their action when they enter the marketplace? Our effort is to identify the factors that are the drivers behind these investment strategies. After several years of work in this area, we are opening up a whole new view of how markets work, together with a new understanding of what’s going on when investors enter the marketplace.

Investment research that is strategy-specific,
not asset-specific

Our first order of business is simply to measure gross flow. How much trading is driven by our $10 trillion sample of professional investment flows? Next, we measure what we consider to be net directional flow - for example, into US equities or the Yen or into German bonds. We divide these flows by the gross flows to give a sense for how big the net is relative to the gross. The third thing we look for is something that doesn’t have the world ?directional? in it -- so there’s an extra absolute value in the top. Instead of averaging over time to see how these flows net out, we’re trying to get an average, for any given day, of how much net flow there is relative to gross flow.

Up until very recently we were working exclusively with equities and foreign exchange flows. Now, with our new Sovereign Bond Flow Indicator, we’re working with government debt as well. And obviously, portfolio flows in one of these asset classes, help to shed light the others - for example the interaction between sovereign debt and foreign exchange

These studies give us a unique perspective on global financial positioning. For example, we can see that US gross equity volumes declined considerably since the year 2000, recovering only recently - but that during the same period the growth rate of ex-US equity volume was just enormous.

These large high level perspectives on investment flows, when netted out over time, can obscure a lot of information. For example, within a massive directional flow, when we disaggregate the information, we can see that there is actually a lot of disagreement going on. Investors some times act in concert, but by and large they have their own independent views and actions. And so a new area of our work consists of dividing our aggregated flows into different sections to see who is participating in the market and how their actions impact the market as a whole.

These days we spend a lot of time thinking about the characteristics of the assets that are to be found in various investment strategies. But asset characteristics are not all that matters. As anyone involved with portfolio optimization knows, the correlation of the assets with the rest of the portfolio is also going to be important in determining demand for the securities. There are fund-wide and portfolio-wide considerations that go beyond the characteristics of the individual assets that are purchased by any given strategy.

What are the asset characteristics we look for? In stocks there are the usual suspects: size, value, growth, as measured by market to book, momentum. Standard factors in a factor model or characteristics in a characteristic model of stock return. We’ve added to this a carry measure in the stock market - something of a value measure, similar to a dividend price ratio. In foreign exchange, value corresponds to the real exchange rate - a measure of overall competitiveness that measures a country’s ability to pay its way in international markets. In the foreign exchange research we also look for hedging activity.

Having identified these characteristics, we then examine the strategies to measure the sensitivity of given strategies to things like larger market cap, more value versus growth, degrees of momentum, the weights of portfolios to the momentum of their assets, to the size, year and capitalization of a stock, to their dividend yield etc. For foreign exchange we compare positions relative to the absolute size of all positions. And we ask similar questions - are investors tilting toward momentum currencies, toward competitive “real exchange rate” currencies, toward carry or toward short-run momentum measures?

At the end of this process, we develop a multi-dimensional box in which, for every strategy, we measure something of their ‘zip code’ with respect to each one of the different characteristics. How much slope do they place on the real exchange rate if they’re a currency fund? How much tilt do they have toward momentum if they’re a stock or currency fund, etc? We then can use descriptors that exist for every strategy. And we can begin, because of these descriptors and tilts, to embark on a new form of investment analysis. For the first time, we have quantifiable inputs with which to analyze ideas that are strategy-specific -- not asset-specific.

Fixed Income Portfolio Reallocation Analysis

Sebastien Page and Jordan Alexiev

Jordan Alexiev
With our Fixed Income Portfolio Reallocation Analysis (PRA) we offer a blueprint for construction of optimized global fixed income investment portfolios that have higher expected information ratios than cap-weighted benchmarks, holding constant the level of absolute risk we are willing to take.

Our work is centered around several innovations in portfolio theory that are unique to State Street Associates regime-dependent risk estimates, scenario-based return estimates, mean-variance-tracking error optimization, within-horizon risk exposure, risk budgets and fundamental value-at-risk (VAR) sensitivities.

We will argue that risk associated with volatility and correlations between fixed income investments are different throughout different interest rate regimes. We will calculate these regime-dependent metrics and employ scenario-based return estimation to derive our expectations for where different indices will be at the end of a one-year horizon.

Building
regime-dependent,
scenario-based
return estimates

Once settled on our inputs, we will employ a very powerful framework called mean-variance tracking error optimization to help us create optimal allocations among these indices by controlling - at the same time ? for absolute risk and for risk relative to benchmarks or peer groups. Having established our optimal allocations, we will try to bridge the gap between the way that investors think and feel about risk, and the way that they measure it, by using within-horizon risk analysis. Finally, we will decompose each optimal portfolio into its components and show how the overall portfolio risk is driven by its parts.

For any optimization process, we need to agree on two estimates - the volatility of each individual investment and the way that investments interact; and second, a set of expected returns. The traditional method, of taking the full sample of historical data and calculating the interdependence between different assets is not the right way when we?re dealing with fixed-income investments. To do so, we would be giving the same level of confidence to periods of high and low interest rates or to periods when the Federal Reserve is increasing or reducing interest rates.

We partition our sample into periods where interest rates are stable and rising vs. periods of falling interest rates and periods with high interest rates vs. periods of low interest rates. We then combine regime specific parameters with various yield curve scenarios to generate return expectations.

In order to be able to price our indices and incorporate our assumptions about the yield curve, we partition our historical sample into various credit spreads, mortgage-backed spreads and spreads of high-yield securities over Treasuries in order to pinpoint which spreads we want to use when coming up with our expectations of total return. These spreads fluctuate widely through various interest rate regimes. Isolating each of our regimes, we can derive our volatility and return assumptions for each of ten fixed income indices and determine which reallocation is most suitable depending on the investor’s relative risk aversion to benchmark tracking error.

• US Government Intermediate

• US Government Long

• US Credit Intermediate

• US Credit Long

• US Mortgage-Backed

• US High Yield

• US TIPS

• Emerging Markets

• Sovereign Bonds

• Global Corporate Bonds

Most investors outperform their absolute target while underperforming their benchmark - or outperform their benchmark while underperforming both their absolute target. We consider this to be a ‘tolerable’ outcome. Clearly, the worst situation is to under-perform the absolute target and the peer group benchmark. We can consider this type of unfortunate investor to be both ‘wrong’ and ‘alone’. What we are after in our effort to develop quantitatively supported optimization frameworks is just the opposite. These lucky, or wise, investors - who outperform both their absolute targets and their peer/benchmark - are in the terrific position of being both ‘right’ and ‘alone’.

To achieve this outcome we employ Mean-Variance-Tracking Error (MVT) optimization. Multi-risk or MVT optimization is an innovation in the field of portfolio theory that maximizes expected return while simultaneously minimizing both absolute variation and deviations from a specified benchmark. This process allows investors to construct portfolio allocations that balance their appetite for risk against their desired tracking error target. This method of portfolio construction - which generates a smooth trade-off between return, risk, and tracking error - is almost always superior to constrained MV optimization and produces a three-dimensional ‘efficient surface’ that incorporates expected return, standard deviation and tracking error.

Sebastien Page
The principal problem facing most fixed income investors is an allocation problem. We want to invest in fixed income indices, incorporating estimates of both return and risk. For return, we use an estimate of total return; to gauge volatility, we use a total return measure that incorporates every aspect of risk. We present three yield curve scenarios, and we look at the volatility of specific interest rate regimes.

In developing our optimal reallocations, we came up with very interesting results. There was no systematic bet or duration shift, nor were there significant allocations to any specific factor. Fixed income portfolio allocation is really a ‘correlation game’. For intermediate maturities we will seek more credit exposure than the benchmark allocation. For longer securities, we’ll seek more Government exposure than credit. When investing globally, we’ll seek more credit than sovereign, versus the benchmark.

In doing all of this, it might seem as though we are tilting our portfolio in favor of volatility. But at the same time, our optimizer is going to overweight the mortgage-backed security asset class, which has only 3% volatility. This is the power of portfolio theory - all these things are happening at the same time and we end up with the exact same volatility and some excess return. There is some peer group risk associated with this excess return but it’s a worthy tradeoff.

In terms of a list of factors, we’re not taking any systematic bet, whether on the credit side or the shape of the yield curve or on the level of interest rates. We hold duration constant. Information ratios are higher for low relative risk portfolios that are closer to the benchmark. If investors don’t pay attention to normal or typical allocations they can opt for portfolios with very high Sharpe ratios - keeping volatility constant and increasing returns. At some point a pure mean-variance portfolio may have uncomfortable weights and therefore be, for practical purposes, not investable.

In our Fixed Income Portfolio Reallocation Analysis, we build regime-dependent risk estimates and scenario-based return expectations. We identify optimal reallocations for investors who care about outperforming a fixed income benchmark while maintaining the same level of absolute risk. We also show the exposure to loss of these reallocations at the end of the horizon as well as throughout the horizon. And we convert these optimal reallocations into risk budgets. We show the impact of each position on portfolio value at risk, as well as the fundamental VAR sensitivities.

State Street Investor Confidence Index

Paul O’Connell, FDO Partners
The American writer Mark Twain once said “all you need in this life is ignorance and confidence, and then success is sure”. As is often the case with Twain’s observations, there’s more to this than meets the eye. Twain neatly identifies the difference between knowledge and confidence - the distinction between fundamentals, and how investors feel about those fundamentals.

In the same way, to create an index of investor sentiment, the goal is to separate market fundamentals - price, earnings, ratios, dividend growth rates etc. - from how investors feel about those fundamentals. This is precisely what State Street’s Investor Confidence Index does. It captures whether investors are bullish or bearish on risk for any given set of fundamentals - whether positive or negative.

There is no guarantee that investor confidence and securities prices will move together. Indeed, it is quite possible that when prices rise, professional investors may decide to sell, thereby taking risk off the table. Investors look to asset prices as a critical element when deciding whether to expand or contract their holdings of risky assets. And that’s precisely the problem in using prices as a proxy for confidence. Prices are an input to confidence, but they are not, of themselves, confidence.

In an effort to more precisely quantify investor confidence, a number of surveys have emerged to gauge investor opinion. Surveys, though, are handicapped by issues of timeliness, reporting accuracy and the inevitable margins of error encountered when extrapolating broad opinion from a narrowly-delivered survey. What’s more, many of the best-known surveys of investor sentiment are directed toward individual - not institutional - investors.

Risk appetite is a
fundamental
building block
for any portfolio

The State Street Investor Confidence Index (ICI) measures confidence directly and objectively by examining the actual investment positions of 10,000 institutional investors managing fully 15% of total world investment. The idea is quite straightforward. The greater the proportion of their portfolio that they deploy across all risky investments, the greater their risk appetite or confidence. Unlike other indices of confidence, the opinions of investors are deemed to be irrelevant - only their actions count. The survey also provides a nice point of contrast to the popular consumer confidence indices that are available.

Risk appetite is a fundamental building block for any portfolio. Every investment decision is composed of three elements: the asset return, the risk associated with the asset and the risk appetite of the investor contemplating the investment. Financial markets are awash with information about the first two and nearly devoid of insight with regard to investor risk appetite.

While the ICI does provide unique insight into investor confidence, it is not a leading market indicator, nor is it necessarily predictive. The routine deviation of the ICI and leading market indicators makes this clear. But understanding risk tolerance is a key input for investors - whether their approaches are quantitative or qualitative. By gaining a better understanding of peer risk tolerance, investors can more coherently gauge their own confidence and appetite for risk.

As a signal measure of investor behavior, the Investor Confidence Index is concerned with what investors are doing - not with what is happening with various asset classes and subsets. The Index, and more finely-focused sub-indices, may reveal the risk appetite of investors from specific geographic regions or from various style approaches - value investors, growth investors, fixed income managers etc.

An Introduction to the Sovereign Bond Flow Indicator (SBFI)

Shafiq Ebrahim and Brian Garvey
The Sovereign Bond Flow Indicator (SBFI), which extends State Street Global Markets’ investor behavior measures beyond foreign exchange and equities to the fixed income arena, gauges the purchases and sales of domestic sovereign bonds by institutional investors. Domestic sovereign bonds are securities issued by national governments, denominated in local currencies, and traded in local markets.

Two types of the indicator - the Country SBFI and the Cross-Border SBFI ? are produced on a daily basis. The former measures aggregate net flows of sovereign bonds, while the latter focuses only on sovereign bond flows by investors whose base currencies differ from the issuer?s local currency. The Country SBFI and the Cross-Border SBFI cover 11 and 13 developed countries, respectively, in North America, Europe, and the Asia-Pacific region. A Eurozone aggregate is also generated for both products.

While the aggregated market capitalization of global equities, at $25 trillion, is some $6.5 trillion greater than the global bond market cap, fixed income flows are about double worldwide equity and foreign direct investment (FDI) flows. Bond flows are at the center of several critical questions that contemporary markets are trying to answer: How sustainable are global imbalances’ Why are U.S. bond yields so persistently low? And when will the U.S. housing sector roll over?

Bond flows are at the center of critical questions facing today’s global markets

Historically, bond flow analysis has relied heavily on the U.S. Treasury International Capital (TIC) reporting system, which collects data for the United States on cross-border portfolio investment flows and positions between U.S. residents (including U.S.-based branches of firms headquartered in other countries) and foreign residents (including offshore branches of U.S. firms). But as this data is reported with a 45-day lag, it is inherently backward-looking.

The SBFI includes money market activity, which TIC data does not, an important factor given that some 15% of foreign holdings of U.S. Treasuries is of less than one year maturity. And the SBFI largely represents institutional investors, while the TIC data may tend to overemphasize the activities of central banks and hedge funds. Central banks tend to be buyers of last resort, and many hedge funds, while based outside the United States, have a U.S. dollar base, which means that they undertake no currency transactions when investing in the U.S. This makes the TIC data less valuable in foreign exchange analysis, something for which the SBFI is uniquely informative.

The SBFI may be used on its own and together with other traditional factors in fundamental and quantitative investment strategies bond, equity, and foreign exchange markets. For example, the Country SBFI is used in bond market interest rate anticipation and yield curve strategies, as well as in tactical asset allocation across bond and equity markets. The Cross-Border SBFI is used in bond market strategies as well as in currency management.

The SBFI accurately represents raw sovereign bond flows while protecting client confidentiality. At the same time, the informative signal component is extracted from the underlying raw flows by applying unique insights into the nature of bond market investor behavior. For example, the SBFI methodology incorporates the fact that investors purchasing fixed rate bonds are really buying packages of future cash flows of various maturities.

Our research has uncovered evidence of negative correlations between the Country SBFI and the Equity Flow Indicator (EFI) - our measure of equity flows - as investors make asset allocation changes within their portfolios. Negative correlations have also been observed between the Cross-Border SBFI and the Foreign Exchange Flow Indicator (FXFI) - our measure of currency flows -- as investors hedge the exchange rate risk of their bond transactions by selling or buying currency forwards. This suggests that fixed income, equity, and currency portfolio managers can use the SBFI together with existing indicators in developing their investment strategies.