Print

Phasing-in costs in the Petroleum Fund

Introduction

Oil revenues in foreign currency are phased into the Petroleum Fund through the sale of foreign currency and purchase of securities. Until the end of June 2004, most of these phasing-in costs were charged to the buffer portfolio, which is part of Norges Bank’s foreign exchange reserves. Starting in the second half of 2004, these costs have been covered by the Petroleum Fund, which reduces the return on the Fund.

Chart 1 shows the transfers to the Petroleum Fund from 1997 to the present. In the initial years, capital was injected every quarter, and the amounts varied considerably. At the start of 2002, the Ministry of Finance decided to increase the phasing-in activity from quarterly to monthly transfers. At the end of December 2004, the cumulative transfers accounted for roughly 90 per cent of the Fund’s increase in value since 1997. The return measured in international currency has increased the value of the Fund by 20 per cent, while the stronger Norwegian krone exchange rate has reduced its value by 10 per cent.

Chart 1: Transfers to the Petroleum Fund. In billions of NOK

The left-hand scale in Chart 2 illustrates how much each transfer constitutes as a percentage of the Fund’s value. The chart shows that the proportion has stabilised at a low level in recent years. This is the result of smaller but more frequent transfers combined with the growth of the Fund’s cumulative value. Since 2002, the monthly transfer has been 1-3 per cent of the Petroleum Fund’s value, but this proportion will decline as the Fund grows in value.

Chart 2: Transfers to the Petroleum Fund as a percentage of its value

When the Norwegian government transfers oil revenues to the Petroleum Fund, trades are carried out whereby foreign currency is sold and foreign securities (equities and fixed income instruments) are purchased. According to the Ministry of Finance’s guidelines for the Petroleum Fund, securities will rarely, if ever, be sold in connection with a regular transfer to the Fund. Therefore, this article will only look at the transaction costs associated with the purchase of securities. These transaction costs have also been discussed in earlier annual reports, including an article in 1999, which provides an introduction to transaction costs, and an article in 2002, which looks at the costs of large equity trades.

Transaction costs

Since transaction costs reduce the return on the Fund, it is vital to keep them to an absolute minimum. Although these costs may be reduced by improving operational efficiency in the markets, even in a perfect trading system there will never be a total absence of transaction costs. We can distinguish between two categories of transaction costs, direct and indirect.

Direct transaction costs are highly visible to all parties to a trade and are relatively simple to measure. The reason is that direct transaction costs are those costs that accrue in addition to the price paid for a security. Indirect transaction costs are far more difficult to measure. Although a number of models have been developed to estimate these costs, none yields entirely satisfactory results.

Direct transaction costs

Direct transaction costs include settlement costs and commissions to brokers and other operators who help to execute the trades. In some countries, including the UK and Ireland, taxes are also levied on securities purchases. The largest component of direct costs is the brokers’ commission. The size of the commission depends on several factors, including the price of the security in question, the volume, the type of trade, the choice of broker, and the market in which the trade takes place. Direct transaction costs are higher for equities than for fixed income instruments. Many analyses assume that the direct costs of purchasing a fixed income instrument are close to zero.

In recent years, stronger competition among market operators and the evolution of ever-more advanced computer trading systems have driven down the commission costs for institutional investors like the Petroleum Fund.

Indirect transaction costs

The indirect costs of a transaction are difficult to measure since they arise when the price actually paid is higher than the market price at the time of the trade. Indirect transaction costs come in three main categories: liquidity costs, market impact, and opportunity costs.

Liquidity costs

Liquidity costs are costs that are incurred due to small trading volumes in the security to be purchased. Low liquidity might, for instance, result from the lack of an open market for the trading and pricing of the security. Investments in securities with low liquidity carry a higher risk than investments in highly liquid securities, partly because it may be difficult to find a counterparty that is willing to buy the security at the time and price at which the seller wants to sell. There is also considerable uncertainty regarding the correct price of such a security.

Market makers are intermediaries whose principal task it is to provide the markets with optimal liquidity by buying and selling securities. The liquidity costs are the compensation they require for the added risk they take on by trading low-volume securities. Most models of liquidity costs start with the difference between the security’s bid price and ask price. Market makers buy at the bid price and sell at the ask price, and the bid-ask spread is an indication of the market makers’ expected profits.

Since the 1970s the reported bid-ask spread has been considered a good measure of the liquidity costs for equities 1). It expresses the difference between bid price and ask price for trades already closed in the market. Simultaneously, the market will contain many open bid and ask orders, including so-called “limit orders”. These will help decide the current market price, but will not be reflected in the reported bid-ask spread until a binding trade is made. In equity markets it is not unusual that settlement takes place at a price within the reported bid-ask spread. This means that the reported bid-ask price spread will provide an exaggerated estimate of liquidity costs. One solution is to calculate the real or effective spread 2). Another is to calculate the absolute difference between actual price and reported mean price, where the mean price is given as the average of the reported bid and ask price. Both of these methods take account of actual market prices, and therefore provide an adjusted estimate for any trades which occur within the recorded spread. A third approach is to use half of the reported spread, which is deemed to produce a lower and more realistic estimate of the liquidity costs for equities.

In fixed income markets, it is not particularly common for settlement to occur within the reported bid-ask spread. The full spread is therefore a more useful estimate of the liquidity costs for fixed income instruments.

Market impact

The effects of market impact are more significant for equities than for fixed income instruments. The costs associated with market impact arise primarily in trades involving large volumes. If a potential buyer is seeking a larger volume than the market maker is willing to sell at the current price, a premium must be paid to settle the trade. This leads to a rise in the market price.

The market price can also change if a rumour or suspicion arises that a high-volume buyer or seller has new information that is not generally available to the market. The purchase of large blocks can spur rumours that the buyer has information about an upcoming price rise, and a sale of large blocks can spur a suspicion that the seller knows something about an approaching price fall. The market price may change regardless of the veracity of the rumours. If there is no new information, the new price will be based on false expectations and it will eventually fall (or rise) back to where it started.

In theory, market impact is the difference between the price paid and the current (unaffected) market price prior to the trade. The unaffected market price is difficult to observe, and it is therefore very difficult to measure the exact costs linked to market impact.

An investor can try to avoid market impact by splitting the trade into several smaller parts. Another approach is to use the “upstairs” market. This is a special market where equal players meet to buy and sell large volumes. The advantage of the upstairs market is that one deals with a discrete counterparty, and the order is not recorded until after it is settled. The disadvantage of the upstairs market is that the process can take time. Additionally, the direct transaction costs may be substantially higher than trades via the normal market channels.

Opportunity costs

If a security that an investor needs to buy is unavailable at the time and/or price desired, the investor has a choice between several alternative strategies, including deferring the trade, purchasing a security with relatively similar characteristics to his first choice (near substitute), or deciding not to trade. The drawback of all these strategies is that they may lead to opportunity costs. These costs are difficult to measure. Opportunity costs arise, for instance, when an investor decides to defer the purchase and the price rises in the meantime, or buys a substitute that does not perform as well as the first choice, or if he chooses not to trade and it turns out that the security in question later rises in value.

A recognised method – implementation shortfall

One means of calculating transaction costs is implementation shortfall. The method, although introduced by Treynor 3)early in the 1980s, is credited to Perold from an article published in 1988 4). The model calculates an aggregate figure for direct and indirect transaction costs. These are found by comparing the actual portfolio with a theoretical portfolio over a time interval.

Before the work on selling foreign currency and buying securities commences, a list is compiled of the securities in which the phasing-in capital will be invested. This theoretical portfolio will consist of all the securities on the list. It is assumed that all purchases in the theoretical portfolio will be executed instantaneously and at the current market price on the decision date, i.e. direct and indirect trading costs will be equal to zero. The actual portfolio will fall short of the theoretical portfolio because, as discussed above, one cannot necessarily buy all the securities on the list at the price and on the date desired. In addition, the actual portfolio will also contain direct trading costs.

Chart 3 shows how the overall transaction costs – known as the implementation shortfall – will be found from the differences in value between the theoretical and the actual portfolio at a preset date or in a preset period.

The method has its appeal because it is simple and straightforward in theory, but unfortunately it is not very simple to apply in practice. For instance, it requires making a clear distinction between the actual portfolio (phasing-in portfolio) and the portfolio that will receive the new capital. This is not the ideal solution when, in addition to estimating the transaction costs, the goal is to ensure the most cost-effective overall management of the Petroleum Fund. Another complication is that the method provides no unambiguous definition of the appropriate decision date, so it is not at all clear when measurement should start and when it should finish.

Market practice

The effort to estimate and measure transaction costs gathered momentum in the late 1990s, but national and international regulations still fall short with regard to rules and guidelines for measurement and presentation of transaction costs. The need for overall guidelines and market standards for the calculation of transaction costs has been discussed recently in articles and consultation papers 5).

The Global Investment Performance Standards (GIPS) is one set of guidelines for the calculation and presentation of investment results which is highly respected by financial market participants all over the world. Work has been under way for many years on a GIPS annex that will deal with transaction costs. Moreover, a proposed set of guidelines from the US Securities and Exchange Commission (SEC) has been circulated for comment, though not yet adopted 6). In Norway, there is no published documentation on this topic. Investment banks, fund managers and other participants acting on their own behalf or on commission for others have now taken the initiative to develop methods and systems for estimating transaction costs. These operators all have a high level of theoretical and practical expertise in the field, which would suggest that their models will provide a sound theoretical foundation, producing reasonable estimates of transaction costs. However, these operators have a direct financial interest in setting market standards. For that reason it would be preferable if the standards are set by regulatory agencies or other bodies that are independent of the market.

Calculation of the implementation shortfall is the method most recognised by market operators. Since it has proven difficult to implement in practice, simplified variants are employed where direct and indirect costs are estimated separately.

Half the bid-ask spread is often employed as an estimate of liquidity costs in equity markets. There is broad agreement as to what factors cause market impact. There is far less agreement as to how these factors should be combined to accurately arrive at a good estimate of the scale of market impact costs. In equity markets, proprietary models are often used to estimate market impact and opportunity costs. Such proprietary models by definition rely on underlying information about the model that is not generally known. In the fixed income markets the greatest focus is on liquidity costs.

In recent years, there has been increasing focus on customer satisfaction, which has highlighted the qualitative aspects of the phasing-in implementation. For instance, it is becoming standard practice for the client and the intermediary who is charged with the task of performing the phase-in to meet in advance to determine an estimate of the transaction costs. Afterwards there is a review of the way the phase-in was handled. The aim here is that the client will gain an understanding of the positive and negative differences between the “actual” transaction costs and the estimate that was given in advance.

When estimating transaction costs, it is necessary to have access to large volumes of reliable data concerning security prices (e.g bid, ask, close and actual transaction prices), and traded volumes at different points in time. Only few years ago this was a major hurdle in the fixed income markets, especially for the less liquid securities, securities traded in the OTC markets, and securities traded in less established markets. New technology has meant that today there is a much better supply of necessary price data. In recent years, many major market operators have cooperated in order to exchange key data on transaction prices and volumes. The result has been a major improvement in the availability of reliable market data.

Calculation of phasing-in costs for the Petroleum Fund

Transaction costs reduce the Petroleum Fund’s return. For that reason one of the goals of Norges Bank is to keep these costs as low as possible. Up until end-June 2004, new capital was transferred to the Petroleum Fund in the form of securities, and the phasing-in costs were charged to the portfolio from which the capital came (the buffer portfolio). Starting in the second half of 2004 new capital has been transferred in cash, which means that phasing-in costs are charged directly to the Petroleum Fund in the form of a lower return.

The increasing interest in transaction costs among market operators and the transition to a new phasing-in regime have been the key reasons why work has been started to calculate these phasing-in costs. At the very least these calculations must meet current market standards and comply with the unwritten rules that constitute the market’s best practice. Until such market standards are available and/or the authorities set requirements for estimating transaction costs, Norges Bank has chosen to employ the methods and models used by comparable institutions.

Furthermore, it is crucial for the method employed to gain acceptance. The model must be intelligible and verifiable, i.e. it must be possible to perform an independent analysis of the estimates. This is vital for providing the degree of transparency necessary for ensuring confidence and credibility in the management of the Petroleum Fund. Therefore, the model must be as simple as possible, yet sufficiently comprehensive to provide a reasonable estimate of the costs that accrue when foreign currency is transferred to the Petroleum Fund.

A final objective has been to aim for equal treatment of the two asset classes in the Petroleum Fund, and among the funds managed by Norges Bank Investment Management. Any differences between asset classes and the funds should be justified by fundamental differences, for instance by how the markets are organised. This will simplify the model, which is an advantage in terms of intelligibility and verifiability. It will provide equal treatment when measuring the results achieved in each asset class and fund.

Section 2.2 of the Annual Report states that the costs of phasing in new capital during the second half of 2004 were estimated at approximately NOK152 million. This estimate was based on the theory and market practices presented in this article. Chart 4 provides a summary of the transfer amounts and phasing-in costs during the second half of 2004. The left-hand scale is the value of the transfer amount and the right-hand scale the magnitude of the costs. No transfers were made to the Petroleum Fund in December 2004.

Chart 4: Monthly transfer amount and phasing-in costs In millions of NOK

While the phasing-in costs will vary with the size of the phasing-in amount, this is not the only variable. An example of other factors is seen in September, when the transfer was higher but the phasing-in costs were lower than in August. The explanation is found in how the capital was invested. In September, almost the entire transfer was invested in fixed income markets, where phasing-in costs are fairly moderate. In August, some two-thirds of the transfer was invested in European equity markets, where both direct and indirect transaction costs are relatively high.

Direct costs

Commission costs, the largest component of the direct costs, depend in part on the type of trade. At Norges Bank Investment Management between 50 and 80 per cent of transfers to the equity portfolio are made in the form of program trades, while the remainder of the equity transfers are made as single-equity trades. Norges Bank Investment Management has estimated the costs of program trades 7) at roughly three basis points of the traded volume and the costs of trading single equities at 2-15 basis points of the traded volume. The calculations for the latter half of 2004 are found by applying a fixed percentage for program trades and single-equity trades, respectively. The tax cost is found by applying the tax rates in the countries where capital was invested.

Indirect costs

Liquidity costs are calculated by using half of the bid-ask spread for equities and the full bid-ask spread for fixed income instruments. The difference in approach is justified by the pricing in the portfolio against which the Petroleum Fund is evaluated (the benchmark portfolio). The equity benchmark portfolio is valued using closing prices, i.e. the prices prevailing at the close of the trading day. Most securities in the fixed income portfolio are priced at the bid price. In the Petroleum Fund most fixed income instruments are bought at the current ask price and most equities at a price between the bid price and the ask price.

The benchmark portfolio does not receive transfers and therefore does not incur the same phasing-in costs as the Petroleum Fund. All else being equal, the return on the Petroleum Fund must therefore be expected be lower than that on the benchmark portfolio. It can be shown that this difference in return can be estimated using the full bid-ask spread for the fixed income portfolio, and half of the bid-ask spread for the equity portfolio.

Costs will also accrue as a result of large transfers (market impact). This cost component is particularly extensive for equities. The models and systems necessary to make such calculations are based on assumptions that are not generally available. It has therefore not been possible to perform a quality check of the results. In its decision to disregard the costs calculated on the basis of such models, Norges Bank Investment Management has attached considerable importance to its need to have quality-assured models. The reason is that since the suppliers of these systems trade securities on their own and others’ account, they may not have the independence necessary to obtain an objective estimate of the transaction costs.

The value of Norges Bank’s calculations is limited by the lack of a satisfactory solution for estimating the effects of market impact in the case of equities. Empirical studies show that market impact represents a significant part of the phasing-in costs. For instance, StockFactsPro estimates a market impact in 2004 of 25 basis points of the transfer amount (see separate box). Since Norges Bank’s estimates for 2004 disregard market impact, they provide a significant underestimate of the actual transaction costs that accrue when petroleum revenues are transferred to the Petroleum Fund.

Example of transfers via the buffer portfolio to the Petroleum Fund in July 2004. Chart A below shows an example of daily transfers to the buffer portfolio for July 2004. Norges Bank determines its daily foreign currency purchases based on Petoro’s estimates of the monthly flow of foreign currency revenues along with the size of the amount to be transferred to the Petroleum Fund at the end of the month. The Ministry of Finance sets the transfer amount. In July, Norges Bank’s daily foreign currency purchases were EUR 35 million, equivalent to NOK 300 million. The curve in the chart below shows how the transfers from the SDFI and Norges Bank increased the value of the buffer portfolio by a total of NOK 16.3 billion during the month. Chart A: Transfers to the buffer portfolio. July 2004. In millions of NOK At the end of July, NOK 14.6 billion was transferred from the buffer portfolio to the Petroleum Fund, NOK 1.9 billion to investments in equity instruments, and NOK12.7 billion to investments in fixed income instruments. The regional breakdown within the equity and fixed income portfolios is shown in Chart B below. In the equity portfolio, capital was only invested in Europe, while in the fixed income portfolio, capital was invested in the Petroleum Fund’s three geographic regions. Chart B: Transfer to the Petroleum Fund July 2004

Costs of equity trades on international stock exchanges The Equities Department of Norges Bank Investment Management has analysed the transaction costs of equity trades on major international stock exchanges. The implementation of trades is a vital part of the investment process, and efficient execution is critical for achieving an excess return. The goal of such analyses is both to obtain an accurate estimate of the actual costs and to make use of data to make the implementation of equity trades more efficient. In this box a method has been used where the cost is given as the difference between actual transaction price and current market price at the time the trade was sent to the trading desk. Direct costs like commission, taxes and stock exchange fees are not included in the analysis, in order to keep the basis for comparison as similar as possible. The calculations take into account many key aspects of the execution of a trade, including the spread between bid and ask prices, decisions taken by the intermediary executing the trade, and any impact the trade might have on the share price. Even though many other factors may also influence the calculations (such as corporate news, general market trends, liquidity and volatility), the method provides a reasonable comparison of costs, and insight into the transaction process since numerous trades are included in the basis of measurement. The chart below shows the transaction costs of single-stock trades (as opposed to program trades) on selected international stock exchanges in 2004. The chart includes stock exchanges where the share of Norges Bank’s total trading volume exceeds 2.5 per cent. Altogether, trades on these exchanges accounted for more than 85 per cent of Norges Bank’s total trading volume in 2004. The costs are estimated in relation to the exchange where costs were lowest (the German XETRA is set at zero) and stated in basis points of the volume traded. This provides an accurate picture of how much it costs to trade on a given exchange relative to the one with the lowest costs. It will be seen from the chart that for Norges Bank Investment Management, the New York Stock Exchange had the highest costs. In this comparison, account has not been taken of stock exchange fees, etc. Of the stock exchanges included here, the London Stock Exchange in particular has high fees. Relative trading costs (basis points)

Notes:
1) Demsetz, H. 1968. "The Cost of Transacting". Quarterly Journal of Economics, vol. 82, no. 1, pp. 33-53.
2) Roll, R. 1984. "A Simple Implicit Measure of the Effective Bid-Ask Spread in an Efficient Market". Journal of Finance, vol. 39, no. 4, pp.1127-39.
3) Treynor, J. "What does it take to win the Trading Game?", Financial Analysts Journal, vol. 37, no. 1, p. 55-60, 1981.
4) Perold, A. "The Implementation Shortfall: Paper versus Reality". Journal of Portfolio Management, vol. 14, no. 3, p. 4-9, 1988.
5) See for example R. Collie, "Performance Standards for Transition Management", The Journal of Performance Measurement, Vol 7, no. 2, Winter 2002/2003, p. 51-59.
6) U.S. Securities and Exchange Commission, "Concept Release: Request for Comments on Measures to Improve Disclosure of Mutual Fund Transaction Costs", File No. S7-29-03, 2003 (http://www.sec.gov/ rules/concept/33-8349.htm). Cf. reply from Barclays Global Investors and ITG Inc. (http://www.sec.gov/rules/concept/s72903.shtml)
7) Program trading means that more than one equity (or a portfolio of equities) is traded simultaneously. The alternative is a single equity trade. According to The Concise Encyclopedia of Economics, the New York Stock Exchange defines a trade as a program trade if it involves 15 or more equities and has a total value exceeding USD 1 million.