Encyclopedia of Finance Part 16 potx

Chapter 29 MARKET EFFICIENCY HYPOTHESIS MELODY LO, University of Southern Mississippi, USA Abstract Market efficiency is one of the most fundamental research topics in both economics and finance. Since Fama (1970) formally introduced the concept of market efficiency, studies have been developed at length to examine issues regarding the efficiency of various financial markets. In this chapter, we review elements, which are at the heart of market efficiency literature: the statistical efficiency market models, joint hypothesis testing problem, and three categor- ies of testing literature. Keywords: market efficiency; security returns; in- formation; autocorrelation; serial correlation (tests); random walk model; (sub)martingale; hy- pothesis testing; (speculative) profits; trading rules; price formation 29.1. Definition The simplest but economically reasonable state- ment of market efficiency hypothesis is that security prices at any time fully reflect all avail- able information to the level in which the profits made based on the information do not exceed the cost of acting on such information. The cost includes the price of acquiring the information and transaction fees. When the price formation in equity market satisfies the statement, market participants cannot earn unusual profits based on the available information. This classical market efficiency definition was formally intro- duced by Fama (1970), and developed at length by researchers in the field. 29.2. The Efficient Market Model Much of work on this line of research is based on an assumption that the condition of market equi- librium can be stated in terms of expected returns. Although there exists diversified expected return theories, they can in general be expressed as fol- lows: E( ^ pp i,tþ1 ) ¼ [1 þ E( ^ rr i,tþ1 jI t )] Â p i,t, (29:1) where E is the expected value operator; p i,t is the price of security i in period t, r i,tþ1 is the one- period rate of return on security i in the period ending at t þ 1, and E(r i,tþ1 jI t ) is the expected rate of return conditional on information (I ) avail- able in period t. Also, variables with hats indicate that they are random variables in period t. The market is said to be efficient, if the actual security prices are identical to their equilibrium expected values expressed in Equation (29.1). In other words, if the actual security price formation fol- lows the market efficiency hypothesis, there would be no expected returns=profits in excess of equilib- rium expected returns. For a single security, this concept can be expressed as follows: E( ^ ZZ i,tþ1 jI t ) ¼ 0, and Z i,tþ1 ¼ r i,tþ1 À E( ^ rr i,tþ1 jI t ), (29:2) where Z i,tþ1 is the return at t þ 1inexcessof the equilibrium expected returns anticip a ted at t. This concept can also apply to the entire security market. Suppose that market participants use infor- mation, I t , to allocate the amount, l i (I t ), of fund s available to each of n security that makes up the entire security market. If the price formation of each of n security f ollows Equa tion (29.2), then t he total excess market value at t þ 1( ^ VV tþ1 ) equals to zero, i.e. E( ^ VV tþ1 jI t ) ¼ X n i¼1 l i (I t )E( ^ ZZ i,tþ1 jI t ) ¼ 0: (29:3) The general efficient market models of Equations (29.2) and (29.3) are the foundations for empirical work in this area. Researchers in the field largely agree that security prices ‘‘fully reflect’’ all available information has a direct implication: successive returns (or price changes) are independent. Conse- quently, researchers tend to conclude market is efficient if there are evidences that demonstrate E( ^ ZZ i,tþ1 jI t ) ¼ 0andZ i,t is uncorrelated with Z i,tþk for any value of k. S imilarly, if E( ^ VV tþ1 jI t ) ¼ 0and V i,t is uncorrelated w ith V i,tþk for any value of k, market is evident to be efficient. Based on efficiency models in Equations (29.2) and (29.3), two special statistical models, submar- tingale and random walk, are closely related to the efficiency empirical literature. The market is said to follow a submartingale when the following con- dition holds: E( ^ ZZ i,tþ1 jI t ) $ 0 for all t and I t : (29:4) The expected returns c onditional on I t is nonnegative and has an important implication on trading rule. This means investors should hold the security once it is bou ght during any future period, because selling it short cannot generate larger returns. More import- antly, if Equation (29.4) holds as equality, the market is said to follow a martingale. Researchers usually conclude that security prices follow ‘‘patterns’’ and market is inefficient when the empirical e vidences are toward rejection of a martingale model. The security prices exhibit the random walk statistical property if not only that the successive returns are independent but also that they are identically distributed. Using f to denote the dens- ity function, the random walk model can be ex- pressed as follows: f (r i,tþ1 jI t ) ¼ f (r i,tþ1 ) for all t and I t : (29:5) The random walk property indicates that the return distributions would repeat themselves. Evidences on random walk property are often considered to be a stronger supportive of market efficiency hy- pothesis than those on (sub)martingale property. 29.3. The Joint Hypothesis Problem The continuing obstacle in this line of empirical literature is that the market efficiency hypothesis per se is not testable. This is because one cannot test market efficiency hypothesis without imposing restrictions on the behavior of expected security returns. For example, the efficiency models of Equations (29.2) and (29.3) are derived based on a joint hypothesis: (i) the market equilibrium re- turns (or prices) are assumed to be some functions of the information set and (ii) the available infor- mation is assumed to be fully utilized by the mar- ket participants to form equilibrium returns, and thereby current security prices. As all empirical tests of market efficiency are tests of a joint hy- pothesis, a rejection of the hypothesis would al- ways lead to two possible inferences: either (i) the assumed market equilibrium model has little abil- ity to capture the security price movements or (ii) the market participants use available information inefficiently. Because the possibility that a bad equilibrium model is assumed to serve as the benchmark can never be ruled out, the precise inferences about the degree of market efficiency remains impossible to identify. 29.4. Three Categories of Testing Literature The empirical work on market efficiency hypoth- esis can be categorized into three groups. First, weak-form tests are concerned with how well past 586 ENCYCLOPEDIA OF FINANCE security returns (and other explanatory variables) predict future returns. Second, semi-strong-form tests focus on the issue of how fast security price responds to publicly available information. Third, strong-form tests examine whether security prices fully reflect private information. 29.4.1. Weak-Form Tests Controversy about market efficiency centers on the weak-form tests. Many results from earlier works on weak-form tests come directly from the submar- tingale expected return model or the random walk literature. In addition, much of the earlier works consider information set as just past historic re- turns (or prices). The most frequently used proced- ure to test the weak form of efficient markets is to examine whether there is statistically significant autocorrelation in security returns using serial cor- relation tests. A pattern of autocorrelation in se- curity returns is interpreted as the possibility that market is inefficient and market participants are irrational, since they do not fully exploit specula- tive opportunities based on the price dependence. The serial correlation tests are tests of a linear relationship between current period’s returns (R t ) and past returns (R tÀ1 ): R t ¼ a 0 þ a 1 R tÀ1 þ « t , (29:6) where R t is the rate of return, usually calculated as the natural logarithm first differences of the trad- ing price (i.e. R t ¼ ln P t À ln P tÀ1 ; P t and P tÀ1 are the trading prices at the end of period t and of period t À1, respectively.), a 0 is the expected re- turn unrelated to previous returns, and a 1 is the size of first-order autocorrelation in the rate of returns. For market efficiency hypothesis to hold, a 1 needs to be statistically indifferent from 0. After conducting serial correlation analysis, Kendall (1953) concluded that market is efficient because weekly changes in 19 indices of British industrial share prices and in spot prices for cotton and wheat exhibit the random walk property. Roberts (1959) notes that similar statistical results can be found when examining weekly changes in Dow Jones Index. (See also Moore, 1962; Godfrey et al., 1964; and Fama, 1965.) Some researchers later argued that the size of serial correlation in returns offers no precise implications on the extent of speculative profits available in the market. They propose that examining the profitability of various trading rules can be a more straightforward meth- odology for efficiency tests. A representative study that adopted this methodology was done by Alex- ander (1961), where he examines the profitability of various trading rules (including the well-known y% filter rule). Despite a positive serial correlation in return series, he also discovers that y% filter rule cannot outperform buy-and-hold rule. He thus concludes that the market is still an efficient one. Similarly, Fama and Blume (1966) find positive dependence in very short-term individual stock price of the Dow Jones Industrial index. Yet, they also suggest that market is efficient because the overall trading costs from any trading rule, aiming to utilize the price dependence to profit, is suffi- ciently large to eliminate the possibility that it would outperform the buy-and-hold rule. In gen- eral, results from earlier work (conducted before the 1970s) provide no evidence against efficient market hypothesis since they all report that the autocorrelations in returns are very close to 0. As more security data becomes available, the post-1970 studies always claim that there is signifi- cant (and substantial) autocorrelation in returns. Lo and MacKinlay (1988) report that there is positive autocorrelation in weekly returns on portfolios of NYSE stocks grouped according to size. In particu- lar, the autocorrelation appears to be stronger for portfolios of small stocks. According to Fisher’s (1966) suggestion, this result could be due to the nonsynchronous trading effect. Conrad and Kaul (1988) investigate weekly returns of size-based port- folios of stocks that trade on both Wednesdays to somehow alleviate the nonsynchronous trading ef- fect. However, as in Lo and MacKinlay (1988), they find positive autocorrelation in returns and that this pattern is stronger for portfolios of small stocks. On another note, the post-1970 weak-form test studies focus on whether variables other than past MARKET EFFICIENCY HYPOTHESIS 587 returns can improve return predictability. Fama and French (1988) use dividend yield to forecast returns on the portfolios of NYSE stock. They find that dividend yield is helpful for return predictabil- ity. On the other hand, Compbell and Shiller (1988) report that earnings=price ratio increases the return predictability. In summary, recent stud- ies suggest that returns are predictable when vari- ables other than past returns are used and the evidences seem to be against the market efficiency hypothesis that was well supported before the 1970s. 29.4.2. Semi-strong-Form Tests Each of the semi-strong-form tests is concerned with the speed of price adjustment to a particular public information event. The event can be macro- economic announcement, companies’ financial reports, or announcement on stock split. The initial work in this line of research was by Fama et al., (1969), in which they studied the speed of price adjustment to the stock-split announcement. Their results show that the informational implications of a stock split are fully reflected in the price of a share at least by the end of the month, or most probably almost immediately after the day of the stock-split announcement. They therefore conclude that the stock market is efficient because the prices respond quite speedily to new public information. Waud (1970) uses residual analysis to study how fast mar- ket reacts to the Federal Reserve Bank’s announce- ment on discount rate changes. The result suggests that market responds rapidly to the interest-rate announcement even when the Federal Reserve Board is merely trying to bring the discount rate in line with other market rates. Ball and Brown (1968) investigate the price reactions to the annual-earn- ings announcement. They conclude that market participants seem to have anticipated most infor- mation by the month’s end, after the annual-earn- ings announcement. These earlier studies (prior to the 1970s), focusing on different events of public announcement, all find supportive evidences of market efficiency hypothesis. Since the 1970s, the semi-strong-form test studies have been developed at length. The usual result is that stock price adjusts within a day of the announcement being made pub- lic. Nowadays, the notation that security markets are semi-strong-form efficient is widely accepted among researchers. 29.4.3. Strong-Form Tests The strong-form tests are concerned with whether prices fully reflect all available information so that no particular group of investors have monopolistic access to some information that can lead to higher expected returns than others. It is understandable that as long as some groups of investors in reality do have monopolistic access to the information, the strong-form market efficiency hypothesis is impossible to hold. In fact, both groups of special- ists, NYSE (see Niederhoffer and Osborne, 1966) and corporate insiders (see Scholes, 1969), have monopolistic access to information, and which has been documented. Since the strong-form effi- ciency model is impossible to satisfy, the main focus in this line of work is to assess if private information leads to abnormal expected returns, and if some investors (with private information) perform better than others because they possess more private information. The most influential work before the 1970s was by Jensen (1968, 1969) where he assessed the performance of 115 mutual funds. Jensen (1968) finds that those mutual funds under examination on average were not able to predict security prices well enough to outperform the buy-and-hold trading rule. Further, there ap- pears no evidence suggesting that individual mu- tual fund performs significantly better than what we expect from random chances. Using Sharpe– Lintner theory (see Sharpe, 1964; Lintner, 1965), Jansen (1969) developed a model to evaluate the performance of portfolios of risk assets. Most im- portantly, he manages to derive a measure of port- folio’s ‘‘efficiency’’. The empirical results show that on average the resources spent by the funds managers to better forecast security prices do not generate larger portfolio returns than what could 588 ENCYCLOPEDIA OF FINANCE have been earned by equivalent risk portfolios selected either by random selection trading rule or by combined investments in market portfolios and government bonds. Jansen further interprets his results that probably mutual fund managers do not have access to private information. These re- sults are clear in line with strong-form market efficiency models because evidence suggests that current security prices have fully reflected the ef- fects of all available information. After the 1970s, there is less of new research examining investors’ access to private information that is not reflected in security prices. Representative studies were done by Henriksson (1984) and Chang and Lewellen (1984). In tests of 116 mutual funds, Henriksson (1984) reports that there is difference between mu- tual fund returns and Sharpe–Lintner market line. Similarly, Chang and Lewellen (1984) note that examination of mutual fund returns show no sup- portive evidence of fund managers’ superior selec- tion abilities. In short, recent studies largely agree to prior literature’s view that investors with private information are unable to outperform a passive investment strategy. Evidences are still in favor of the existence of market efficiency hypothesis. 29.5. Conclusion This review has been brief and so various issues related to market efficient model have not been considered. Volatility tests of market efficiency, and cross-sectional return predictability based on various asset pricing models are just some of the omitted issues. For more details, readers are referred to two excellent market efficiency survey papers by Fama (1970, 1991). REFERENCES Alexander, S.S. (1961). ‘‘Price movements in speculative markets: trends or random walks.’’ Industrial Man- agement Review, 2: 7–26. Ball, R. and Brown, P. (1968). ‘‘An empirical evaluation of accounting income numbers.’’ Journal of Account- ing Research, 6: 159–178. Chang E.C., and Lewellen, W.G. (1984). ‘‘Market tim- ing and mutual fund investment performance.’’ Jour- nal of Business, 57: 57–72 . Compbell J.Y. and Shiller, R. (1988). ‘‘Stock prices, earnings and expected dividends.’’ Journal of Fi- nance, 43: 661–676. Conrad, J. and Kaul, G. (1988). ‘‘Time-variation in expected returns.’’ Journal of Business, 61(4): 409–425. Fama, E.F. (1965). ‘‘The behavior of stock market price.’’ Journal of Business, 38(1): 34–105. Fama, E.F. (1970). ‘‘Efficient capital markets: a review of theory and empirical work.’’ Journal of Finance, 25(2): 383–417. Fama, E.F. (1991). ‘‘Efficient capital markets: II.’’ Jour- nal of Finance, 46(5): 1575–1617. Fama, E.F. and Blume, M. (1966). ‘‘Filter rules and stock market trading profits.’’ Journal of Business (Special Supplement), 39: 226–241. Fama, E.F. and French, K.R. (1988). ‘‘Dividend yields and expected stock returns.’’ Journal of Financial Economics, 22: 3–25. Fama, E.F., Fisher, L., Jensen, M.C., and Roll, R. (1969). ‘‘The Adjustment of Stock Prices to New Informa- tion.’’ International Economic Review,5:1–21 Fisher, L. (1966). ‘‘Some new stock-market indexes.’’ Journal of Business, 39(1), Part 2: 191–225. Godfrey, M.D., Grange r, C.W.J., and Morgenstern, O. (1964). ‘‘The random walk hypothesis of stock mar- ket behavior.’’ Kyklos, 17: 1–30. Henriksson, R.T. (1984). ‘‘Market timing and mutual fund performance: an empirical investigation.’’ Jour- nal of Business, 57: 73–96 . Jensen, M.C. (1968). ‘‘The performance of mutual funds in the period 1945–64.’’ Journal of Finance, 23: 389–416. Jensen, M.C. (1969). ‘‘Risk, the pricing of capital assets, and the evaluation of investment portfolios.’’ Journal of Business, 42: 167–247. Kendall, M.G. (1953). ‘‘The analysis of economic time- series, Part I: Prices.’’ J ournal of the Royal Statistical Society, 96 (Part I): 11–25. Lintner, J. (1965). ‘‘Security prices, risk, and maximal gains from diversification.’’ Journal of Finance, 20: 587–615. Lo, A.W. and MacKinlay, A.C. (1988). ‘‘Stock market prices do not follow random walks: evidence from a simple specification test.’’ Review of Financial Stud- ies, 1(1): 41–66. Moore, A. (1962). ‘‘A Statistical Analysis of Common Stock Prices.’’ PhD thesis, Graduate School of Busi- ness, University of Chicago. MARKET EFFICIENCY HYPOTHESIS 589 Niederhoffer, V. and Osborne, M.F.M. (1966). ‘‘Market making and reversal on the stock exchange.’’ Journal of the American Statistical Association, 61: 897–916. Roberts, H.V. (1959). ‘‘Stock market ‘patterns’ and financial analysis: methodological suggestions.’’ Journal of Finance, 14: 1–10. Scholes, M. (1969). ‘‘A test of the competitive hypoth- esis: the market for new issues and secondary offer- ing.’’ PhD thesis, Graduate School of Business, University of Chicago. Sharpe, W.F. (1964). ‘‘Capital assets prices: a theory of market equilibrium under conditions of risk.’’ Journal of Finance, 19: 425–442. Waud, R.N. (1970). ‘‘Public interpretation of federal discount rate changes: evidence on the ‘Announce- ment Effect’.’’ Econometrica, 38: 231–250. 590 ENCYCLOPEDIA OF FINANCE Chapter 30 THE MICROSTRUCTURE= MICRO-FINANCE APPROACH TO EXCHANGE RATES MELODY LO, University of Southern Mississippi, USA Abstract The vast empirical failure of standard macro ex- change rate determination models in explaining exchange rate movements motivates the development of microstructure approach to exchange rates in the 1990s. The microstructure approach of incorporating ‘‘order flow’’ in empirical models has gained consid- erable popularity in recent years, since its superior performance to macro exchange rate models in explaining exchange rate behavior. It is shown that order flow can explain about 60 percent of exchange rate movements versus 10 percent at most in standard exchange rate empirical models. As the microstruc- ture approach to exchange rates is an active ongoing research area, this chapter briefly discusses key concepts that constitute the approach. Keywords: microstructure approach; order flow; exchange rates; macroexchange rate models; het- erogeneous information; private information; asset market approach; goods market approach; cur- rency; divergent mappings; transaction 30.1. Definition The microstructure approach to exchange rates is considered to be a fairly new but active research area. This line of research emerged in the early 1990s mostly due to the vast empirical failure of standard macro exchange rate determination models. In more recent years (the late 1990s), there was considerably a large amount of pub- lished work regarding the microstructure approach to exchange rates, suggesting order flow is evident to be the missing piece in explaining exchange rate behavior. The following definition of the micro- structure approach to exchange rates comes dir- ectly from its pioneer, Richard Lyons (See Lyons, 2001). The microstructure approach is a new approach to exchange rates whose foundations lie in micro- economics (drawing particularly from microstruc- ture finance). The focus of the approach is dispersed information and how information of this type is aggregated in the marketplace. By dispersed information, we mean dispersed bits of information about changing variables like money demands, risk preferences, and future inflation. Dispersed information also includes information about the actions of others (e.g. about different trading responses to commonly observed data). The fact that the private sector might be solv- ing a problem of dispersed information is not con- sidered in traditional macro models. Rather, macro models assume that information about vari- ables like money demands, risk preferences, and inflation is either symmetric economy-wide, or in some models, asymmetrically assigned to a single player – the central bank. In reality, there are many types of dispersed information that ex- change rates need to impound. Understanding the nature of this information problem and how it is solved is the essence of this micro-based research agenda. 30.2. Empirical Failure of Traditional Approaches To Exchange Rates The literature has documented extensively the little ability traditional=standard exchange rate deter- mination models have to explain exchange rate behavior. Meese and Rogoff (1983) show that a random walk model outperforms the standard international-finance models in forecasting ex- change rates. In that respect, Meese (1990) writes that ‘‘ . . . the proportion of (monthly or quarterly) exchange rate changes that current models can explain is essentially zero . . . This result is quite surprising, since exchange rate changes would be entirely unpredictable only in very special cases of the theoretical models discussed.’’ More recently, a survey paper by Frankel and Rose (1995) also notes that ‘‘To repeat a central fact of life, there is remarkably little evidence that macroeconomic variables have consistent strong effects on floating exchange rate, except during extraordinary circum- stances such as hyperinflations.’’ Two most frequently discussed standard exchange rate determination approaches are (1) goods market approach and (2) asset market approach. The goods market approach suggests that exchange rates move to reflect necessary changes in excess demand=supply of foreign cur- rency resulting from international trades. A do- mestic economy necessarily demands for more foreign currencies when its citizens consume more imported goods. The general prediction of goods market approach is that an increase in domestic trade deficit must lead to the depreciation of do- mestic currency against foreign currency. How- ever, existing studies find no empirical evidence to support any specific relation between current ac- count imbalance and exchange rate movements. In open economies, domestic citizens can pur- chase not only foreign goods but also foreign fi- nancial assets. The asset market approach suggests that demand for foreign currency increases when domestic citizens increase their possessions on for- eign assets, and this in turn would cause domestic currency to depreciate against foreign currency. Different from the goods market approach, the asset market approach also concerns the market efficiency issue. Specifically, the theoretical models on asset market approach determine equilibrium exchange rate at the level that no public informa- tion can lead to excess returns. In general, the empirical model specification for asset market approach is as follows (Lyons, 2001): DE t ¼ f 1 (i,m,z) þ « 1t , (30:1) where DE t is changes in nominal exchange rate (usually monthly or weekly data is used), the func- tion f 1 (i ,m,z ) includes the current and past values of domestic and foreign interest rates (i), money sup- ply (m), and all other macro variables (z). Similar to the low predictability of goods market approach, the majority of asset market empirical studies re- port that macro variables in Equation (30.1) explain 10 percent only, at most, of exchange rate move- ments. Further details on the empirical failure of various standard exchange rate determination models are well documented by Taylor (1995). The disappointing results from the existing ex- change rate models motivated researchers to look for sources responsible for the empirical failure. They attribute the general empirical failure to the unrealistic assumptions shared among standard exchange rate determination models. In detail, these models assume that every market participant learns new information at the same time when macroeconomic information=news is made public. Further, all market participants are assumed to have the ability to impound macro information into prices to the same level. However, both as- sumptions can easily be argued. In reality, not only 592 ENCYCLOPEDIA OF FINANCE market participants’ information set is heteroge- neous, but also their mapping ability from avail- able information to price is impossible to be the same. The heterogeneity in information set is evi- dent from the fact that foreign exchange traders, working for different banks, each have their own customers to deal with. Transactions with different customers offer each trader ‘‘private’’ information that he may not intend to share with others. In addition, it is understandable that different people tend to interpret the market impact of new in- formation on exchange rate differently, regardless whether the information is made available to all of them at the same time. This idea of divergent mappings from information to prices is discussed by Isard (1995, pp. 182–183) who states that ‘‘econo- mist’s very limited information about the relation- ship between equilibrium exchange rates and macroeconomic fundamentals, . . . it is hardly con- ceivable that rational market participants with complete information about macroeconomic fun- damentals could use that information to form pre- cise expectations about the future market-clearing level of exchange rates.’’ 30.3. Why Microstructure Approach? The unrealistic assumptions in standard exchange rate models mentioned above have been relaxed in the literature that aims to explain why the financial market crashed. It is important to note that despite events such as stock market crash and currency crisis appear to be macro issues, they can be largely explained by microstructure approach that con- siders the existence of heterogeneous information among market participants (see Grossman, 1988; Romer, 1993; Carrera, 1999). For the same token, Lyons argues that adopting microstructure ap- proach to investigate the trading process of ex- change rates may help our understanding on when and how exchange rates move. Lyons (2001, p. 4) notes that the microstructure approach is an approach that relaxes three of the assets approach’s most uncomfortable assumptions. First, on the aspect of information, microstructure models recognize that some information relevant to exchange rates is not publicly available. Second, on the aspect of players, microstructure models recognize that market participants differ in ways that affect prices. Last, on the institutional aspect, microstructure models recognize that trading mechanism differs in ways that affect prices. 30.4. The Information Role of Order Flow The central variable that takes the fundamental role in microstructure approach, but has never been presented in any of previous exchange rate models, is order flow. Order flow is cumulative flow of signed transaction volume. A simple ex- ample on how order flow is counted for individual transaction can be helpful. Suppose that a dealer decides to sell 5 units of U.S. dollars via a market order (one unit usually represents a transaction worth $1 million), then order flow is counted as – 5. The negative sign is assigned for this $5 million transaction because it is a seller-initiated order. Each transaction is signed positively or negatively depending on whether the initiator of the transac- tion is buying or selling. Over time, order flow gives us a relative number of buyer-initiated versus seller-initiated orders in a market. Thus, order flow provides information to dealers about the relative demand for currencies at any time in the market. Since market participant must make buy-or-sell decisions according to available infor- mation (including their private information), it is presumed that order flow is at certain level driven by market fundamentals. Order flow plays a fundamental role in exchange rate movements because it has the function to transmit information that is not known by every- one in the market. In fact, this concept of order flow transmitting information is intuitionally appealing. As an example to describe the intuition, consider two traders (referred to dealer A and dealer B) in the foreign exchange market, and each of them trades for a particular bank. Each bank of course has its own customers from whom it buys and sells foreign exchange. When THE MICROSTRUCTURE=MICRO-FINANCE APPROACH TO EXCHANGE RATES 593 dealer A trades with his own customers, he obtains private information, such as the customers’ view of the current market (price), and which, is not known to dealer B. However, when dealer A puts orders in the inter-dealer market in an attempt to balance out positions with outside customers (for inventory concern), dealer A’s private information is learned by dealer B. An alternative example is related to the idea of divergent mappings from (public) information to prices. Suppose dealer A hears a macro announcement at the same time as dealer B. Although they do not know how each other would interpret the announcement’s effect on prices, they can learn this information by watching how each other trades. A related question that is frequently asked is ‘‘does order flow really contain (market) informa- tion?’’ The answer is positive. The direct evidences come from dealers themselves. In surveys con- ducted by Cheung and Wong (2000), about 50 percent of dealers who responded to the survey claim that they believe banks with larger customer base have information advantage. This is because they get to trade with more customers, and more transactions ensure more private information, which leads to better speculative opportunities. Further evidence is from empirical analysis, which examine whether order flows have a per- manent effect on prices. The rationale behind this empirical analysis is if order flow does not contain any information about market fundamentals, it can only have transitory effect on prices. French and Roll (1986) have used this methodology to identify the information arrival. Using vector auto- regression models, Evans (2001) and Payne (1999) found that order flow innovation has long-run effect on prices. This result provides evidence that order flow does contain information related to market fundamentals. The general empirical model specification for microstructure approach to exchange rates can be written as follows (Lyons, 2001): DE t ¼ f 2 (X,I,Z) þ « 2t , (30:2) where DE t is changes in nominal exchange rate between two transactions, function f 2 (X,I,Z) in- cludes the order flow (X), dealers’ inventory (I), and all other micro variables (Z). The microstruc- ture models predict that an upward move in price is associated with a situation in which buyer- initiated trades exceed seller-initiated trades. In other words, to support microstructure approach to exchange rate, there needs to be a positive rela- tion between order flows and prices. Lyons (2001) and Evans and Lyons (2002) have shown the con- siderably strong positive impact of order flow on exchange rates. More precisely, they have shown that order flow can explain about 60 percent (ver- sus 10 percent at most in standard exchange rate empirical models expressed in Equation (30.1) ) of exchange rate movement. 30.5. Conclusion The high explanatory power of order flow for exchange rate movements is exciting news for re- searchers in the area. So far, all empirical evidences have suggested order flow is indeed the important missing piece in exchange rate determination. Lyons (2001) thus claims that order flows help solve three exchange rate puzzles: (1) the determin- ation puzzle, (2) the excess volatility puzzle, and (3) the forward-bias puzzle. Yet, there is not much agreement toward this claim (see Dominguez, 2003). Clearly, more research needs to be done before these puzzles may be solved. REFERENCES Carrera, J.M. (1999). ‘‘Speculative attacks to currency target zones: a market microstructure approach.’’ Journal of Empirical Finance, 6: 555–582. Cheung, Y.W. and Wong, C.Y.P. (2000). ‘‘A survey market practitioners’ view on exchange rate dynam- ics.’’ Journal of International Economics, 51: 401–423. Dominguez, K.M.E. (2003). ‘‘Book Review: Richard K. Lyons , The microstructure approach to exchange rates, 2001, MIT Press’’ Journal of International Eco- nomics, 61: 467–471. 594 ENCYCLOPEDIA OF FINANCE [...]... and Oxford: Elsevier, North-Holland, p 3, 168 9–1729 French, K.R and Roll, R (1986) ‘‘Stock return variances: the arrival of information and the reaction of traders.’’ Journal of Financial Economics, 17: 5–26 Grossman, S.J (1988) ‘‘An analysis of the implications for stock and futures price volatility of program trading and dynamic hedging strategies.’’ Journal of Business, 61: 275–298 595 Isard, P (1995)... MICROSTRUCTURE=MICRO -FINANCE APPROACH TO EXCHANGE RATES Evans, M (2001) ‘‘FX trading and exchange rate dynamics.’’ Working Paper No 8 116, National Bureau of Economic Research Evans, M and Lyons, R.K (2002) ‘‘Order flow and exchange rate dynamics.’’ Journal of Political Economy, 110: 170–180 Frankel, J and Rose, A (1995) ‘‘Empirical research on nominal exchange rates,’’ in Handbook of International Economics... Fluctuations in the PostBretton Woods Era.’’ The Journal of Economic Perspectives, 4(1): 117–134 Meese, R and Rogoff, K (1983) ‘‘Empirical exchange rate models of the seventies.’’ Journal of International Economics, 14: 3–24 Payne, R (1999) ‘‘Informed trade in spot foreign exchange markets: an empirical investigation.’’ Typescript, London School of Economics Romer, D (1993) ‘‘Rational asset-price movements... investigation.’’ Typescript, London School of Economics Romer, D (1993) ‘‘Rational asset-price movements without news.’’ American Economic Review, 83: 1112–1130 Taylor, M.P (1995) ‘‘The economics of exchange rates.’’ Journal of Economic Literature, 83: 13–47 . of investment portfolios.’’ Journal of Business, 42: 167 –247. Kendall, M.G. (1953). ‘‘The analysis of economic time- series, Part I: Prices.’’ J ournal of the Royal Statistical Society, 96 (Part. Jour- nal of Business, 57: 73–96 . Jensen, M.C. (1968). ‘‘The performance of mutual funds in the period 1945–64.’’ Journal of Finance, 23: 389– 416. Jensen, M.C. (1969). ‘‘Risk, the pricing of capital. the size of serial correlation in returns offers no precise implications on the extent of speculative profits available in the market. They propose that examining the profitability of various trading