Foreign Exchange Market Efficiency Slides to highlight: Course web - PowerPoint PPT Presentation

First, many researchers have tried to measure the profits earned by market professionals such as mutual fund managers. If these managers achieve superior returns (after adjustment for risk) then the market is not efficient with respect to the information processed by the managers. This approach has the advantage that it concentrates on real trading by real market participants, but it has the disadvantage that one cannot directly observe the information used by the managers in their trading strategies. As an alternative, one can ask whether hypothetical trading based on an explicitly specified information set would earn superior returns. To implement this approach, one must first choose an information set. The classic taxonomy of information sets distinguishes among Weak- form Efficiency, Semistrong-Form Efficiency, and Strong- Form Efficiency. 7-31

Foreign Exchange Market Efficiency ~ r Define as the actual one-period rate of  j , t 1 return on asset j in the period ending at time t+1 , ~ ( | ) E r I and as the expected return  , 1 j t t conditional on available information ( I ) at time t . Then the excess market return ( Z ) can be ~   written as: ( | ) Z r E r I    j , t 1 j , t 1 j , t 1 t 7-32

Foreign Exchange Market Efficiency An efficient market has two defining characteristics: Firstly, the expected excess market return, ( | ) E Z I should equal 0, and secondly,  , 1 j t t z z should be uncorrelated with  , , j t j t k for any value of k. That is,    ( ) ( ) ( ) E Z Z E Z E Z   , , , , j t j t k j t j t k for any value of k. z z are not Note that and  , , j t j t k necessarily independent. 7-33

Foreign Exchange Market Efficiency These two properties together implies that the sequence { Z t } is a fair game with respect to I t . In words, the market is efficient if, on average, errors in the formulation of expectations about prices or returns are zero, and these errors follow no pattern that might be exploited to produce profits. 7-34

Foreign Exchange Market Efficiency At times, we think about efficiency in terms of the level of prices instead of the rate of return for convenience. The link between today’s price ( P t ) and the expected future price E ( P t+1 | I t ) is given by: ~   E ( P | I ) [ 1 E ( r | I )] P   1 1 t t t t t ~ ( | ) E r I where is the expected equilibrium  t 1 t yield on spot market speculation. 7-35

Foreign Exchange Market Efficiency Again, market efficiency requires that the sequence of expected errors (X) follow a fair- game process ~   ( | ) X P E P I    1 1 1 t t t t 7-36

Pictures of Efficient Markets When Equilibrium Expected Returns are Constant When prices evolve as a random walk, then tomorrow’s price ( P t +1 ) is equal to today’s price ( P t ) augmented by an error term ( u t +1 ). The distribution of the error term ( u ) is independent and identically distributed over time. 7-37

Efficient Market Behavior with a Constant Equilibrium Expected Return r t . . .   . . . . . . . . . . . ~ . . . . . . . . . . . . . E r I r 0 . . . . .  . . . . 1 . t t . . . Time 7-38 Figure 7.1

Efficient Market Behavior with a Constant Equilibrium Expected Return    ( ) r u  P P e 0 1 t We can write this as:  1 t t Taking the natural logarithm, we have    ln( ) ln( ) P P r u   1 0 1 t t t  0  prices follow a random walk without drift r 0  0  prices follow a random walk with drift r 0 7-39

7-40

7-41

7-42 Figure A: Exchange Rate Levels

7-43 Figure B: Exchange Rate Percentage Changes

Each dot shows the returns on the New York * Composite Index on two successive weeks between January 1968 and January 2007. The circled dot shows a weekly return of 3.4%, followed by 4.9% in the next week. The scatter diagram shows no significant relationship between returns on successive weeks. New York Composite Index Weekly Percentage Changes Source: Brealey, Myers and Marcus, Corporate Finance 7-44

Each dot shows the returns on the New York Composite Index on two successive months between January 1968 and January 2007. This scatter diagram shows that there is also no relationship between market returns in successive months. New York Composite Index Monthly Percentage Changes Source: Brealey, Myers and Marcus, Corporate Finance 7-45

If you are not sure what we mean by “random walk,” consider the following example: You are given $100 to play a game. At the end of each week a coin is tossed. If it comes up heads, you win 3% of your investment; if it is tails, you lose 2.5%. Therefore, your payoff at the end of the first week is either $103 or $97.50. At the end of the second week the coin is tossed again. Now the possible outcomes are as follows: This process is a random walk because successive changes in the value of your stake are independent. That is, the odds of making money each week are the same, regardless of the value at the start of the week or the pattern of heads or tails in the previous weeks. If a stock’s price follows a random walk, the odds of an increase or decrease during any day, month, or year do not depend at all on the stock’s previous price moves. The historical path of prices gives no useful information about the future—just as a long series of recorded heads and tails gives no information about the next toss. Source: Brealey, Myers and Marcus, Corporate Finance 7-46

Can you tell which is which? The top chart shows the real Standard & Poor’s Index for the years 1980 through 1984. The bottom chart was generated by a series of random numbers. You may be among the 50% of our readers who guess right, but we bet it was just a guess. Source: Brealey, Myers and Marcus, Corporate Finance 7-47

Cycles self-destruct as soon as they are recognized by investors. The stock price instantaneously jumps to the present value of the expected future price. Source: Brealey, Myers and Marcus, Corporate Finance 7-48

Efficient Market Behavior with a Constant Equilibrium Expected Return    ( ) r u  P P e 0 1 t We can write this as:  1 t t Taking the natural logarithm, we have    ln( ) ln( ) P P r u   1 0 1 t t t  0  prices follow a random walk without drift r 0  0  prices follow a random walk with drift r 0 7-49

Efficient Market Behavior with a Constant Equilibrium Expected Return Recall our discussion of the International Fisher Effect (IFE), where the future spot exchange rate ( S t +1 ) is modeled as the current spot rate ( S t ) adjusted by the return differential on the two currencies. ~   ( ) i i E S S     $ £ 1 t t i i  $ £ 1 S i £ t 7-50

Efficient Market Behavior with a Constant Equilibrium Expected Return If we augment the IFE with an error term ( u ), we   have:     ( ) i i u  S S e $ £ 1 t  1 t t Taking the natural logarithm, we have:     ln( ) ln( ) ( ) S S i i u   1 $ £ 1 t t t The above equation portrays the spot exchange rate as following a random walk with drift equal to the interest differential. 7-51

Pictures of Efficient Markets When Equilibrium Expected Returns Wander Substantially Efficient market behavior continues to require that the actual returns oscillate randomly about expected returns to meet the criterion of a fair game. However, in this case it is clear that the underlying asset prices did not evolve as a random walk with zero drift, or a random walk with constant drift, or a random walk with any other obvious pattern of deterministic drift. 7-52

Efficient Market Behavior when the Equilibrium Expected Rate of Return Wanders Substantially   ~ E r I  . 1 t t . . . . . r t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . r 0 . . . .. . . . . . . . . . . . . . . . . . . . Time 7-53 Figure 7.2

Efficient Market Behavior when the Equilibrium Expected Rate of Return Wanders Substantially In the sticky-price version of the monetary model, we saw that in response to an unanticipated increase in the domestic money supply, the exchange rate depreciates immediately by an amount greater than what is required in the long run, and then appreciates asymptotically back to its long-run equilibrium value. 7-54

Efficient Market Behavior when the Equilibrium Expected Rate of Return Wanders Substantially During the adjustment period, exchange rate changes are serially correlated and efficient market behavior requires that the actual exchange rates oscillate randomly about the benchmark. Again, because the interest differential along the adjustment path always equals the percentage exchange rate change, there are no profit opportunities even though the adjustment path exhibits a trend. 7-55

Interpreting Efficient Market Studies In Figure 7.1, the series {r t } appears to be priced ~  efficiently against the benchmark E ( r | I ) r  t 1 t 0 But it would be priced inefficiently versus any other choice. Similarly, in Figure 7.2, the series {r t } appears to be priced efficiently against the ~ benchmark E ( r | I )  t 1 t But it is priced inefficiently against the ~  E ( r | I ) r benchmark  t 1 t 0 7-56

7-57

7-58

Interpreting Efficient Market Studies This illustrates that all tests of market efficiency are tests of a joint hypothesis :- (1) the hypothesis that defines market equilibrium prices or market equilibrium returns as some function of the available information set, and (2) the hypothesis that market participants have actually set prices or returns to conform to their expected values. 7-59

Interpreting Efficient Market Studies For empirical studies that reject market efficiency, it is impossible to determine whether an incorrect specification of the market’s equilibrium benchmark is responsible for the rejection or whether market participants were indeed inefficient information processors. The theory of exchange rate determination developed in Chapter 6 found that various exchange rate levels and paths of adjustment could be offered as equilibrium paths. 7-60

Interpreting Efficient Market Studies It need not be the case that the equilibrium exchange rate takes on a constant value, or that it follows a simple linear trend, or some other deterministic pattern. The theoretical criterion of efficiency is for exchange rates to deviate randomly and with mean zero from their equilibrium value, which we have argued could itself wander substantially and in a serially correlated fashion. 7-61

Defining the Available Information Set It is common to distinguish three types of market efficiency depending upon the information set, I t . These are (Fama (1970)):  Weak form, in which the current price reflects all information in the historic series or prices.  Semistrong form, in which the current price reflects all publicly available information.  Strong form, in which the current price reflects virtually all available information, including proprietary and insider information. 7-62

Information and the Levels of Market Efficiency Strong Form: All Public and Private Information Semistrong Form: All Public Information Weak Form: Past Prices 7-63 Investments

Fama (1991) proposed the following taxonomy:  Tests of return predictability; indicating studies that examine whether returns can be predicted by historic prices or historic information on fundamental variables.  Event studies, referring to studies that examine how prices respond to public announcements.  Tests for private information, including studies that examine whether specific investors have information not in market prices. Fama argues that the new terminology is more descriptive of the empirical work and consistent with the common usage. 7-64

Weak-Form Tests and Tests of Predictive Ability Given our discussion of equilibrium or benchmark models, it is apparent that weak-form tests of market efficiency (or tests of predictive ability) must be formulated and interpreted with caution. Specifically, a test of whether the exchange rate follows a random walk or some other time series process cannot be offered as a test of efficiency when divorced from a model of the equilibrium exchange rate. 7-65

Weak-Form Tests and Tests of Predictive Ability Analysis of the statistical properties of exchange rates, however, may be useful for descriptive purposes. For example, measuring deviations from PPP is useful for assessing changes in competitiveness across countries. 7-66

Semi-Strong Form Tests and Event Studies Similarly, semistrong form tests that draw on publicly available information (such as forward exchange rates and interest rates) will be heavily dependent on the model of equilibrium. Monetary models of the exchange rate assume that financial assets denominated in different currencies are perfect substitutes. 7-67

Semi-Strong Form Tests and Event Studies With this assumption, we showed in Chapter 6 that the interest differential between domestic and foreign assets should equal the anticipated exchange rate change, and that the forward premium is an unbiased forecaster of the future exchange rate change. 7-68

Semi-Strong Form Tests and Event Studies However, within the class of portfolio balance models, financial assets denominated in different currencies are imperfect substitutes. According to this benchmark, the forward premium is a biased forecaster of the anticipated exchange rate change as a result of an exchange risk premium. Clearly, we must agree upon the exchange rate model before we can interpret semistrong form tests of market efficiency. 7-69

Semi-Strong Form Tests and Event Studies Empirical tests of the role of news are event studies that illustrate similar joint-hypothesis testing problems. In response to news about the money supply, interest rates, the fiscal budget deficit, and so on, we showed in Chapter 6 that a currency might logically appreciate or depreciate depending on the scenario for the future, which is typically unknown at the time of news release. 7-70

Strong-Form Tests and Private Information Strong form tests of market efficiency examine whether market prices fully reflect information available only to market insiders. This information set could include knowledge of intervention in the market by central bankers that is often kept secret, knowledge of customer orders that is available to interbank market makers, and proprietary models of exchange rate forecasting that have not been published or made available to a wide audience. 7-71

Market Efficiency with Uncertainty and Risk Investment When the future spot rate is a random variable, the investor who holds a net (asset or liability) position in foreign currency is exposed to foreign exchange risk. Because a test of market efficiency tests a joint hypothesis, the specification of the expected equilibrium return for bearing exchange risk is critical. However, there is no general agreement on the appropriate model for the equilibrium pricing of foreign exchange risk. So, tests of efficiency under uncertainty will not lead to definitive results. 7-72

There are basically two techniques for bearing exchange risk: spot speculation and forward speculation. In either case, the profit depends on the expected future spot rate E ( S I ) which is uncertain.  t 1 t  1 i  F S When interest parity holds, t t *  1 i And spot and forward speculation are equivalent investments that produce the same expected profits. Institutional factors and transaction costs will lead investors to pick the spot or forward market as the preferred venue for speculation. 7-73

The long run is a misleading guide to current affairs. In the long run we are all dead. Economists set themselves too easy, too useless a task if in tempestuous seasons they can only tell us when the storm is long past, the ocean will be flat. - John Maynard Keynes 7-74

Spot Market Efficiency: Design of the Tests The primary technique for testing spot market efficiency has been to compute the profitability of various mechanical, or technical strategies. Most of these studies are examples of weak-form tests (or tests of return predictability) that use only the past series of exchange rates to generate buy and sell trading signals. 7-75

Spot Market Efficiency: Design of the Tests A filter rule is defined by a single parameter (f), the filter size. An f percent filter rule identifies trends and generates buy and sell signals according to the following design: Buy a currency whenever it rises f percent above its most recent trough; Sell the currency and take a short position whenever the currency falls f percent below its most recent peak. Typically, f is chosen to be a small number (e.g., 1%). 7-76

Mechanics of a Filter Rule in the Foreign Exchange Market Time series graph of $/DM exchange rate $/DM t 1 t 2 t 3 t 4 t 5 t 6 t 7 t 8 t 9 t 10 Time 7-77 Figure 7.3 (top)

At the start of the process (t 1 ), the speculator has no foreign exchange positions. But she does have capital that earns interest at the risk-free rate and allows the speculator to enter into the transaction that follow. Note: An f-percent filter rule generates buy signals when the currency rises f-percent above an interim trough (at points t 2 , t 6 , and t 10 ), and sell signals when the currency falls f-percent below an interim peak (at points t 4 and t 8 ). Initially (at t 1 ), the speculator has a net worth that allows him to execute transactions, but he holds no foreign exchange positions. 7-78

After a buy signal, the speculator takes a long position in DM by borrowing US$ and buying DM in the spot market. After a sell signal, the speculator sells his DM holdings and takes a short position in DM [of the assumed size of US$1,000] by borrowing DM and buying US$ in the spot market. The above transactions and positions are described in the following T-accounts. 7-79

Mechanics of a Filter Rule in the Foreign Exchange Market Speculative trading position over time t 1 t 2 t 4 t 6 t 8 0 0 DM $ $ DM DM $ $ DM Net worth ($) 7-80 Figure 7.3 (bottom)

At time t 2 , the DM is assumed to have risen by 1%. The filter rule signals an upward trend in the DM, and so the trading strategy calls for a spot DM ($) purchase (sale). The speculator borrows an amount of US$ and uses it to purchase DM spot, which is invested in an interest-bearing account. The speculator’s position (long DM and short US$) is described by the T-account corresponding to time t 2 . 7-81

The cost of taking on this position is the interest differential (i t2,$ - i t2,DM ). The cost of holding this position for m days is  t m 2   [( i i )] t ,$ t , DM t 2 7-82

Compounding and Continuous Compounding A 0 - principal; r - interest rate; t - the time period to maturity; A - principal and interest     t A A 1 r If interest is calculated 0 once every period nt    r    A A 1 If interest is calculated n 0   n times every period nt    r    lim A A 1 If interest is compounded 0   n   continuously n mrt    1    Let m = n / r lim 1 A A 0   m   m 7-83

Compounding and Continuous Compounding n    r    lim 1 e   n   n Since  where 2 . 7182818284 59045 e So  rt A A e 0 For example, if $1,000 is invested for 1 year at a nominal rate of 10% compounded continuously, the future value at the end of that year is given as follows:     0 . 10  0 . 10 1   A $ 1 , 000 e $ 1 , 000 2 . 71828 $ 1 , 105 . 17 So the effective rate is 0.10517. 7-84

Compounding and Continuous Compounding If the $1,000 is invested at a nominal rate of 10% for 3 years, the future value, assuming continuous compounding, is equal to:     0 . 30  0 . 10 3   A $ 1 , 000 e $ 1 , 000 2 . 71828 $ 1 , 349 . 86 The effective rate is $ 1 , 349 . 86   3 1 . 34986 1 . 10517 3 $ 1 , 000 Again, the effective rate is 0.10517. 7-85

Compounding and Continuous Compounding To generalize, the effective rate is calculated as r   nominal effective rate e 1 r   nominal 1 effective rate e        e r ln 1 effective rate ln r nominal nominal 7-86

e = 2.7 1828 1828 45 90 45 e x = 1 + x + x 2 /2! + x 3 /3! + x 4 /4! + ... lim (1 + r/k) kt = e rt k->  lim (1 + 1/x) x = e (we let x = k/r = kt so that rt = 1) k->  The expected return on the currency position is ~ ( ) E S ~ ~     t m ln ln ( ) ln R 2 E S S  t m t S 2 2 t 2 7-87

Continuous Compounding and Logarithmic Returns In Interest Rate Parity, we balance the return on a US$ investment with the covered return on a US$investment with the covered return on a UK £ investment. Let’s assume that the interest rates on the two securities are in continuous terms, the arbitrage condition becomes: 7-88

Continuous Compounding and Logarithmic Returns Interest Rate Parity Conditions 1 $ i i      $ £ $ 1 e $ 1 e F ( ) $ £ S ( ) £ $ F ( ) i $ e  £ i i   $ £ e $ i £ e S ( ) £ F   ln i i $ £ S 7-89

Continuous Compounding and Logarithmic Returns International Fisher Effect condition ~ 1 i i      $ 1 $ $ 1 £ ( ) e e E S  t 1 $ S ( ) £ ~ i E ( S ) $ e   i i t 1   e $ £ i S e £ t ~ ( ) E S  t 1   ln i i $ £ S 7-90

Continuous Compounding and Logarithmic Returns Using continuous returns rather than simple returns is especially helpful when analyzing a time series of prices or returns. Suppose that over N periods, a security appreciates from price P 0 to price P N . The total price change (P N -P 0 ) can be decomposed into a sequence of prices changes (P N -P N-1 ) + (P N-1 -P N-2 ) + …+ (P 2 -P 1 ) + (P 1 -P 0 ) 7-91

The logarithmic returns over these N periods: ln(P N /P N-1 ), ln (P N-1 /P N-2 ), …, ln(P 2 /P 1 ), ln(P 1 /P 0 ), when added together yield the total return ln(P N /P 0 ). The simple mean and standard deviation of logarithmic returns result in unbiased estimates of average return and volatility. 7-92

Return Definitions and Conventions Prices, Returns, and Compounding Let P t be the price of an asset at date t and assume that this asset pays no dividends. The simple net return , R t , on the asset between dates t -1 and t is : P   t 1 R t P  1 t The simple gross return on the asset is 1 + R t . The asset’s gross return over the most recent k periods from date t-k to date t , written 1+ R t ( k ), is equal to the product of the k single-period returns from t-k +1 to t :                  1 1 1 1 R k R R R    1 1 t t t t k 7-93 Campbell : The Econometrics of Financial Markets pgs 9-11

P P P P P            1 2 1 t t t t k t P P P P P      1 2 3 t t t t k t k The asset’s net return over the most recent k periods, written R t ( k ), is equal to its k -period gross return minus one. These multiperiod returns are called compound returns. Multiyear returns are often annualized to make investments with different horizons comparable : 1 / k        1 k         Annualized 1 1 R k R  t t j    0 j 7-94 Campbell : The Econometrics of Financial Markets pgs 9-11

Since single-period returns are generally small in magnitude, the following approximation based on a first-order Taylor expansion is often used to annualize multiyear returns :  1 k   1     Annualized R k R  t t j k  0 j The difficulty of manipulating geometric averages motivates another approach to compound returns. The continuously compounded return or log return r t of an asset is defined to be the natural logarithm of its gross return (1+ R t ) :   P      t log 1 log r R p p  1 t t t t P  1 t  log p P 7-95 t t

The advantages of continuously compounded returns become clear when we consider multiperiod returns :         log 1 r k R k t t                 log 1 1 1 R R R    1 1 t t t k                 log 1 log 1 log 1 R R R    1 1 t t t k        r r r    1 1 t t t k Hence, the continuously compounded multiperiod return is simply the sum of continuously compounded single-period returns. 7-96 Campbell : The Econometrics of Financial Markets pgs 9-11

The expected profit from using the filter rule strategy at time t 2 is  ~ ~   E ( ) R C t 2 By time t 3 , the DM has hit its peak. But the filter rule does not signal a change till time t 4 . At time t 4 , a sell signal causes the speculator to sell the original DM position (using the proceeds to repay the US$ loan) and short the DM (at S t4 ) [of the assumed size of US$1,000] in anticipation of a further fall in its price. The speculator’s new position (long US$ and short DM) is shown by the T-account corresponding to time t 4 . 7-97

The cost of taking on this position is the interest differential (i t4,DM - i t4,$ ). The cost of holding this  t n 4   [( )] i i position for n days is t , DM t ,$ t 4 The expected return on the currency position is  S _ Re ~ ~ Short Sell venue     t ln ln ln ln ( ) R 4 S E S ~   t t n _ Short Sell Cost ( ) 4 4 E S  t n 4 ~  E ( S ) S This return is positive as  t n t 4 4 which is the expected price at which the speculator will cover the short position. 7-98

The speculator’s profit (  ) represents the incremental return from accepting a foreign exchange risk. Recall that the speculator has pledged a certain amount of capital (net worth) which allows her access to the borrowing and lending capabilities of the foreign exchange market. The speculator’s capital is assumed to earn interest at a competitive market rate (R F ). Thus, the total return from committing this capital to a trading strategy is R F +  . 7-99

Currency Futures Markets Let’s apply the method to the currency futures markets. In this case, the graph would represent the $/DM price of a DM currency futures contract. A long DM position implies buying the DM futures, and a short DM position implies selling the DM futures. It would be necessary to take into account the interest cost of any short position or interest earned on any long position because the futures price itself already reflects these interest rates through the interest rate parity condition. 7-100