American depository receipts, listing, and market efficiency: three case studies.

Introduction

American Depository Receipts (ADRs) are negotiable certificates that represents a foreign company's publicly traded equity (Cooper, 1997). ADRs are created when a foreign company's shares are purchased abroad and delivered to a depository's local custodian bank, usually

Owners of ADRs have a legal claim on the cash flows of the deposited shares. The depository bank receives the dividends, which are then paid to the holder of the ADR less a small handling fee. In 1995, more than 350 firms were listed on U.S. stock exchanges, either directly or indirectly (through ADRs). Although many foreign companies are content to allow their shares to be traded in the form of ADRs, others choose to directly list their shares on a major U.S. stock exchange, such as the New York Stock Exchange (NYSE) or the American Stock Exchange (AMEX). This decision partly reflects the company's belief that direct listing will make it easier to raise capital in the U.S. market by maximizing corporate visibility, market efficiency, and liquidity.

Trading ADRs

ADRs may be traded like any other security, either on an organized stock exchange or in the NASDAQ over-the-counter market. ADRs may also be used to raise capital. The common stock of many of the world's most well known multinational companies arc traded as ADRs, including Nestle (Switzerland), Volvo (Sweden), and British Gas (United Kingdom). ADRs can be resold to another resold in the foreign market.

ADRs offer a number of benefits to investors seeking to diversify internationally. ADRs greatly facilitate trading in foreign securities by reducing the risk of fraud. While foreign companies shares typically are written in the language of the issuer, ADRs are usually issued in the language of the issuing agent. ADRs are legal obligations of the issuing agent and not of the firm that issued the stock. Thus, the risk of falling prey of bogus certificates is eliminated. As such, ADRs overcome many of the obstacles that mutual funds, pension funds, and other financial institutions have in investing and holding securities outside the U.S.(GEIDRS 1997).

ADRs are also convenient. Securities do not have to be delivered through international mail, prices are quoted in U.S. dollars, and pay dividends or interest in U.S. dollars. In fact, the prices of a number of foreign stocks routinely are reported in the American financial press. Importantly, global custodian safekeeping charges associated with purchasing foreign securities are eliminated, which could save the investor as much as 40 basis points annually.

An important function of ADRs is that they enable foreign firms to raise capital in the United States, which is the world's preeminent market for investment capital. Listing shares directly on U.S. stock exchanges, however, is problematic, Disclosure requirements are among the strictest in the world. Foreign firms also face significant costs producing U.S.-style financial statements. ADRs provide foreign firms with a way around these listing problems.

Types of ADRs

Foreign companies can choose from four types of ADR facilities: Unsponsored and three levels of sponsored ADRs. Unsponsored ADRs are issued by one or more depository banks in response to market demand. There is, however, no formal agreement between the depository bank and the foreign company. Unsponsored ADRs are considered obsolete and are no longer created because of a lack of control and hidden transaction costs. Sponsored ADRs, on the other hand, are issued by a depository bank appointed by the foreign company under a deposit agreement or service contract. Sponsored ADRs provide a measure of control over the facility, flexibility to list on a U.S. stock exchange, and the ability to raise capital.

There are three levels of sponsored ADRs. A sponsored Level I ADRs is the simplest method for foreign companies to access U.S. and non-U.S. capital markets. Level 1 ADRs, which are traded over-the-counter and on some foreign stock exchanges, enable foreign companies to establish a foothold into the U.S. equities market by building a core of domestic investors at relatively low cost. Moreover, foreign company do not have to comply with U.S. Generally Accepted Accounting Principles (GAAP) or full Securities and Exchange Commission (SEC) disclosure. Level I ADRs are relatively illiquid, but they also do not permit foreign companies to raise capital, or to be listed on a national exchange.

Foreign companies that wish to formally list their securities on a major U.S. exchange, or raise capital in U.S. markets, must utilize sponsored Level II or Level III ADRs. While each of these two levels require adherence to GAAP, they differ in SEC registration and reporting requirements. Companies that utilize Level II or Level III ADRs must also satisfy the listing requirements of the exchange (GEIDRS, 1997).

Level II ADRs may be traded on an organized stock exchange, such as NYSE or AMEX, or in the NASDAQ over-the-counter market. Level II ADRs are publicly listed, but do not include registered public offerings at the time of listing. The main advantages of Level II ADRs are that they offer greater investor visibility and, presumably, increased market liquidity. The primary disadvantages are that foreign companies must adhere to full SEC disclosure requirements and incur continuous reporting costs.

Level III ADRs, which may also be traded on an organized stock exchange in the NASDAQ over-the-counter market, may be sold in a public offering. As in the case of Level Il ADRs, Level III ADRs must also adhere to GAAP, comply with full SEC disclosure requirements, and incur continuous reporting costs (GEIDRS, 1997).(1)

Market Efficiency and Liquidity

Although many ADRs are considered highly liquid financial assets, many depository banks argue that directly listing shares on major exchanges, such as the New York Stock Exchange, carries with it benefits that outweigh the costs of complying strict disclosure U.S. reporting requirements, including greater investor interest, and a broader and more diversified investor base, which implies increased market efficiency and liquidity.

Liquidity refers to the speed and ease with which an asset can be converted into cash. A necessary condition for market efficiency is liquidity. Markets are said to be efficient if current market prices fully, immediately, and without bias reflect all publicly available information. Institutional rigidities and lags in the information transmission mechanism, however, suggests that market efficiency is functionally related to the size of the market and transaction volume. Greater investor interest and broader market participation suggest that new information will be more rapidly disseminated and reflected in share prices.

The manner in which exogenous shocks are incorporated into share prices also is related to the quality and quantity of publicly available information. When new information is accurate and reliable, market prices should adjust monotonically with a relatively short lag. Where information is less accurate, or where the information transmission mechanism is less efficient, share prices may overreact to new information. In such cases, share prices may "overshoot," but eventually converge to the correct price. This correction may be monotonic or may exhibit improper oscillations. In either case, the greater the market efficiency the more rapidly will share prices converge to a steady state following an exogenous shock.

One important implication of perfectly efficient markets is that investments in those markets have zero net present value - investments are neither undervalued nor overvalued. When markets are efficient, investors get exactly what they pay for when they buy securities and firms receive exactly what their stocks and bonds are worth when they sell them.

The purpose of this paper is to test the hypothesis that the secondary market for foreign shares traded directly on major U.S. exchanges is more efficient than when these shares are indirectly traded through ADRs by applying the Dickey-Fuller unit-root test to the daily closing prices of Telecomunicacoes Brasileiras, S.A. (Telebras), Astra, A.B., and Mavesa, S.A. shares as ADRs and as shares directly listed on NYSE. The paper also examines structural change in the market for foreign equities by examining the dynamic properties of share prices traded as ADRs and as listed securities.

Although the selection of Telebras, Astra, and Mavesa was somewhat arbitrary, this was made necessary by the fact that there are currently over 1,600 depository receipt programs for companies from over 60 countries. The selection criteria included the requirements that each company be recently listed directly on NYSE, that each represent a diverse selection of countries, and that they represent a wide range of industries. Telebras (Brazil), Astra (Sweden), and Mavesa (Venezuela) shares were directly listed on NYSE in 1995, 1996, and 1997, respectively. Telebras, Astra, and Mavesa are leading international companies in the telecommunications, pharmaceuticals, and consumer household products industries, respectively.

Review of the Literature

There is a substantial body of literature on the relationship between asset yield and market efficiency. Much of the literature identifies three levels of market efficiency: Weak, semi-strong, and strong. Each of these levels relates to specific assumptions regarding the information set available to investors and the manner in which market prices are established (Fama, 1970).

A market is characterized as weakly efficient if contemporaneous securities prices fully reflect the information implied by all prior price movements. Current price movements are totally independent of previous price movements, which would imply the absence of price regularities with prophetic significance. Such price variations have been referred to as "random walks." Nelson and Plosser (1982) have found evidence that many macroeconomic time series follow random walks. Their work spawned a number of other studies that investigated whether economic and financial time series are random walks or whether they tended to revert to some long-run trend following an exogenous shock, such as an unforeseen economic, political, or social event. Several of these studies have argued that many economic time series are random walks, or have random walk components.

Weak-level efficiency implies that there is no linear correlation between current monthly returns and returns in prior months. Markets that are weakly efficient may be described by the expression

(1) [P.sub.t] - [P.sub.t-1] = E([R.sub.t]) + [[Epsilon].sub.t]

and

(2) [R.sub.t] = {[P.sub.t] + [I.sub.t] - [P.sub.t-1]}/[P.sub.t-1]

where [R.sub.t] is the rate of return in period t, [P.sub.t] is the weighted average security price, [P.sub.t-I] is a weighted average security price in period t - 1, [I.sub.t] is the interest payment during period t, and [.sub.t] is a random error term.

Equation (1) is sometimes referred to as a "random walk with drift." This period's price is equal to the previous period's observed price, plus the expected return on the security (the drift) and a random error term (the random walk). The expected return on the security is assumed to respond instantaneously and without bias to newly published information. The random error can be either positive or negative, although its expected value is zero. The distinguishing characteristic of markets that are weakly efficient is that they rule out the possibility of trading rules that exploit correlated regularities in past price movements to generate above-average returns.

Data

Telecomunicacoes Brasileiras, S.A. is Brazil's primary supplier of public telecommunications services. Through a network of 27 subsidiaries, Telebras provides local phone service to about 12.1 million customers, and cellular phone service to another 1.3 million customers. Through another subsidiary, Empresa Brasileira de Telecomunicacoes, S.A., Telebras also provides all of Brazil's interstate and international long-distance telephone services. Either directly or through its various subsidiaries, Telebras also is a provider of data transmission, image, videotext, telex and telegraph services.

Sweden's Astra, A.B. is an international pharmaceuticals company specializing gastrointestinal, respiratory, cardiovascular, pain control, central nervous system, and anti-infective products. Astra also produces a line of medical devices. Astra ranks as one of the world's largest pharmaceutical companies with about 20,000 employees, 67 percent of whom are located outside Sweden. Astra markets its products through agents and licensees in more than 40 countries. The company's research and development activities primarily are conducted by five subsidiaries in Sweden and in the United Kingdom. In 1996 while the international pharmaceutical market grew by about 7 percent Astra's sales increased by about 9 percent (17 percent at constant exchange rates).

Mavesa, S.A. is one of Venezuela's largest manufacturers, marketers and distributors of branded consumer processed foods and cleaning products, including margarine, mayonnaise and other related food products, laundry soaps and soap by-products. Mavesa also has an oil palm plantation and a shrimp farm. In the fiscal year ended October 31, 1996 Mavesa's products held the largest share of Venezuela's retail margarine (82 percent), mayonnaise (55 percent), laundry soap (76 percent), vinegar (56 percent), ketchup (20 percent) and baby food (25 percent) markets. On November 29, 1995, Mavesa acquired the assets and brands of Yukery Venezolana de Alimentos, S.A., a leading Venezuelan producer of generic food products, including fruit juices, fruit preserves, ketchup, tomato sauce, baby food, chocolate drinks and mineral water. Yukery's annual sales represent approximately 10 percent of Mavesa's annual sales.

To test the hypothesis that the secondary market for foreign shares traded directly on major U.S. exchanges are more efficient that when these shares are indirectly traded through ADRs this study examined daily closing prices for Telebras for the period November 22, 1993 to December 31, 1996, Astra for the period September 26, 1994 to June 30, 1997, and Mavesa for the period December 1, 1994 to October 31, 1997.

At the time of listing on NYSE, the shares of each of these companies were traded as Level II ADRs (no public offering at the time of listing). The selection of sampling periods was dictated by the NYSE listing date of each of the issues traded and the availability of data. In general, this study examined approximately two years of daily closing prices for ADRs prior to formal listing on NYSE, and approximately one year of daily closing prices after listing.

Telebras was traded in the form of American Depository Receipts until November 1, 1995 when it was formally listed on NYSE. At the time of listing the Telebras conversion rate was one ADR per 1,000 preferred shares. Astra was traded in the form of ADRs until May 23, 1996 when it was formally listed on NYSE. At the time of listing the Astra conversion rate was one ADR per one series "A? ordinary share. Finally, Mavesa shares were traded as ADRs until January 6, 1997 it was formally listed on NYSE. At that time the Mavesa conversion rate was one ADR per 40 ordinary shares.

Statistical Tests

There is a large body of literature investigating weak-level market efficiency and random walks using unit-root tests introduced by Dickey and Fuller (1979, 1981) and Fuller (1976). To understand the rationale behind the Dickey-Fuller unit-root test for weak-level efficiency, assume that share prices over time may be described by the expression

(3) [P.sub.t] = [Alpha] + [Beta]t + [Rho][P.sub.t-1] + [u.sub.t]

where t is a time index. For [Beta] [greater than] 0, [Rho] [less than] l, [P.sub.t] has a positive trend that is stationary after de-trending.

Another possibility is where [Alpha] [greater than] 0, [Beta] = 0, [Rho] = 0. In this case [P.sub.t] is said to follow a random walk with a positive drift. In this case, it would be desirable to work with [Delta][P.sub.t] since de-trending will not result in a stationary time series, and inclusion of [P.sub.t] in a regression (even if de-trended) could lead to spurious results. Moreover, it is not possible to test the statistical significance of estimated [Rho] using a t-test since if the true value of p is equal to unity then ordinary-least-squares estimators will be biased toward zero, which could lead to a rejection of the random walk hypothesis. Dickey and Fuller (1981) derived the distribution of critical F-values ([F.sup.*]) to test the random walk hypothesis i.e., [H.sub.0]: [Beta] = 0 and [Rho] = 1.

Suppose that [P.sub.t] is described by the equation

(4) [P.sub.t] = [Alpha] + [Beta]t + [Rho][P.sub.t-1] + [Lambda][Delta][P.sub.t-1] + [[Epsilon].sub.t]

where ?[P.sub.t-1] = [P.sub.t-1] - [P.sub.t-1]. Using the method of ordinary-least-squares, Equations (5) and (6) are estimated as the unrestricted and restricted equations, respectively.

(5) [P.sub.t] - [P.sub.t-1=] [Alpha] + [Beta]t + ([Rho]-1)[P.sub.t-1] + [Lambda][Delta][P.sub.t-1]

(6) [P.sub.t] - [P.sub.t-1] = [Alpha] + [Lambda][Delta][P.sub.t-1]

A Wald F-test is conducted to test the null hypothesis that [Beta] = 0 and [Rho] = 1.(2)

Table 1 summarizes ordinary-least-squares parameter estimates and Wald F-statistics ([F.sub.w]) for Equations (5) and (6) for the rate of return on investments in Telebras, Astra, and Mavesa shares during the entire sampling period. The numbers in parenthesis are estimated standard errors. Also included in the table are the respective error sums of squares (ESS) for both the unrestricted and restricted equations, from which were calculated Wald F-values. When [F.sub.w] [less than] [F.sup.*] we may not reject the hypothesis is a random walk.(3)

[TABULAR DATA FOR TABLE 1 OMITTED]

The results presented in Table 1 suggest that for the entire sampling period the secondary markets for Telebras, Astra, and Mavesa shares, either indirectly as ADRs or directly as equity on NYSE, were weakly efficient. This result is consistent with the theoretical and empirical literature regarding equity share prices on well developed stock exchanges. The purpose of this study is to examine whether the secondary markets for Telebras, Astra, and Mavesa shares were significantly structurally different in the periods before and after directly listing on NYSE.

Structural Change

Market efficiency implies that new information is quickly and without bias incorporated into share prices. This section examines whether the secondary market for Telebras, Astra, and Mavesa shares became more efficient following formal listing on NYSE, which would imply that new information was more rapidly incorporated into share prices after listing. In each case the entire sampling period was divided into two sub-sampling periods based upon the company's NYSE listing date. In the case of Telebras, the first and second sub-sampling period were November 23, 1993 to October 31, 1995 and November 1, 1995 to December 31, 1997, respectively. In the case of Astra, the first and second sub-sampling periods were September 26, 1994 to May 22, 1996 and May 23, 1996 to June 30, 1997, respectively. Finally, in the case of Mavesa, the first and second sub-sampling periods were December 1, 1994 to January 6, 1997 and January 7, 1997 to October 31, 1997, respectively.

Table 2 summarizes the ordinary-least-squares parameter estimates and [F.sub.w] values for Equations (5) and (6) for Telebras, Astra, and Mavesa ADRs during the first sub-sampling periods. A comparison with the critical values at traditional confidence levels summarized in Dickey and Fuller (1981) indicate that in each case [F.sub.w] [less than] [F.sup.*], which suggest that the secondary markets for Telebras, Astra, and Mavesa ADRs were weakly efficient.

Table 3 summarizes the parameter estimates and [F.sub.w] values for Equations (5) and (6) for Telebras, Astra, and Mavesa traded directly on NYSE during the second sub-sampling periods. Once again, [F.sub.w] [less than] [F.sup.*] suggests that the secondary markets for Telebras, Astra, and Mavesa shares can be characterized as weakly efficient.

[TABULAR DATA FOR TABLE 2 OMITTED]

[TABULAR DATA FOR TABLE 3 OMITTED]

The empirical results presented in Tables 2 and 3 indicate that the secondary market for Telebras, Astra, and Mavesa shares was weakly efficient in each sub-sampling periods examined. Within the context of market efficiency and liquidity, in the cases examined were there any marginal benefits derived from direct listing on NYSE?

A complete cost-benefit analysis associated with directly listing on a major stock exchange is beyond the scope of this study. The public relations and other benefits from directly listing on NYSE have been well documented, as are the high costs of satisfying NYSE listing requirements. Nevertheless, the issues of increased market efficiency and liquidity are important in the decision to directly list because it affects the ability of a company to raise new capital at an optimal price in equity markets.

Although it has been shown that the secondary markets for Telebras, Astra, and Mavesa were weakly efficient in both sub-sampling periods, did these markets experienced structural change following direct listing on NYSE? To test the hypothesis of structural change a Chow test was conducted using the data presented in Tables 2 and 3.(4) This statistic is used to test the null hypothesis of structural stability against the alternative hypothesis of no structural stability.

If the calculated Chow F-statistic is greater than the critical F-value ([F.sup.**])then we reject the null hypothesis in favor of its alternative. The value [F.sup.**] at 2 and 120 degrees of freedom at the 95 percent and 99 percent confidence levels are 3.07 and 4.79, respectively. These results are presented in Table 4.

Table 4. Chow Test for Structural Stability

Company           [F.sub.c]

Telebras            0.96
Astra               1.11
Mavesa              0.57

The results presented in Table 4 indicate that at traditional confidence levels there was no structural change between the first and second sampling period. These results verify that the secondary markets for Telebras, Astra, and Mavesa were weakly efficient in both before and after direct listing on NYSE. Do these results suggest that there were no benefits realized in terms of increased market efficiency and liquidity from directly listing on NYSE? To examine this question more closely, a consideration of the dynamic properties of the restricted price relationships are considered in the next section.

Market Efficiency and Convergence

To examine the potential benefits of increased market efficiency and liquidity from directly listing Telebras, Astra, and Mavesa shares on the NYSE consider the dynamic properties of the restricted price relationships summarized in Tables 2 to 4 by rewriting Equation (6) as

(7) [P.sub.t] + [a.sub.1][P.sub.t-1] + [a.sub.2][P.sub.t-2] = g(t)

where [a.sub.2] = ?, and g(t) = ?. Equation (7) is the general form of a second-order difference equation (Baumol 1970; Gandolfo 1971). The general solution to the homogeneous (left-hand side) part of Equation (7) may be written as

(8) [[Lambda].sup.2] + [a.sub.1][Lambda] + [a.sub.2] = 0.

Equation (8) is called the characteristic (or auxiliary) equation of the homogeneous part of Equation (7). The roots of this equation are given as

[[Lambda].sub.1,2] = {-[a.sub.1] [+ or -] [-square root of [Omega]]}/2

where [Mathematical Expression Omitted]

The dynamic properties of secondary market share prices depends on the value of the sign of the discriminant ?. With the above restrictions on [a.sub.1] and [a.sub.2] the general solution to the homogeneous part of Equation (7) becomes

(9) [P.sub.t] = [A.sub.1] + [A.sub.2] [[Lambda].sup.t]

where [A.sub.1] and [A.sub.2] are two arbitrary constants. If the absolute value of), is less than unity then the movement of [P.sub.t] over time will be convergent on [A.sub.1]. If the absolute value of [Lambda] is greater than unity then the movement of [P.sub.t] over time will be divergent from [A.sub.1]. In general, divergence is not expected since this would suggest market instability, which is not observed in reality. The movement of [P.sub.t] over time also will be monotonic or exhibit improper oscillations (alternating between positive and negative) depending on the signs of [Lambda].

If [Lambda] is positive and less than one then [P.sub.t] will monotonically converge to [A.sub.1]. On the other hand, if [Lambda] is negative and less than one in absolute value then the movement of [P.sub.t] will be convergent, but will exhibit improper oscillations, suggesting that investors overreacted to new information. In such cases, share prices are said to "overshoot," but eventually converge to the correct price.

An examination of the parameter estimates of the restricted equations in Tables 2 and 3 suggest that the secondary market for Telebras and Astra shares became more efficient in the period following direct listing on NYSE. In each case, the parameter estimates in Table 3 indicates that [Delta][P.sub.t-1] was statistically insignificant, which implies that new information was incorporated virtually instantaneously into Telebras and Astra share prices in the period immediately following the direct listing on NYSE.

An examination of the parameter estimates presented in Table 2 suggest that while the secondary market for Telebras and Astra ADRs could also be characterized as weakly efficient new information was incorporated into share prices somewhat less rapidly than after direct listing. The estimated value of value of [Lambda] for Telebras and Astra was statistically significant at the 76 and 96 percent confidence levels, respectively, which suggest that new information was incorporated into Telebras and Astra ADR prices in about 20 and 25 hours, respectively.(5)

In the case of Telebras and Astra, convergence to a new ADR price was monotonic, which suggests that the market correctly evaluated the impact of the new information (there was no "overshooting"), although there was brief delay in processing the information, probably because of somewhat smaller market participation than after direct listing. These results suggest that from the point of view of market efficiency and liquidity that Telebras and Astra benefitted from directly listing its shares on NYSE and opposed to continued trading in ADRs.

Finally, the empirical results suggest that in the case of Mavesa market efficiency and liquidity was unaffected by NYSE listing. The value of [Delta][P.sub.t-1] was statistically insignificant both before and after listing, which indicate that new information was instantaneously and without bias incorporated into Mavesa share prices both before and after listing. This would suggest that there were few efficiency and liquidity gains from direct listing on NYSE

Summary

This paper examined the weak-level efficiency of the secondary market for Telecomunicacoes Brasileiras, S.A. (Brazil), Astra, A.B. (Sweden) and Mavesa, S.A. (Venezuela) shares as ADRs and as shares directly listed on NYSE by applying the Dickey-Fuller unit-root test, and by examining the dynamic properties of the resulting estimated equations. This study finds that the secondary market for these companies' shares can be characterized as weakly efficient both before and after listing on NYSE.

In the period before direct listing of Telebras and Astra shares on NYSE, new information was monotonically incorporated into share prices in about 20 and 25 hours, respectively, suggesting that the market correctly evaluated the impact of the new information, albeit with a slight delay. This may have been the result of lower public interest in ADRs, and therefore somewhat smaller market participation, than in directly listed shares, which would suggest that ADRs were marginally less liquid. After direct listing on NYSE, new information was incorporated instantaneously into Telebras and Astra share prices, which suggests that there were marginal efficiency and liquidity gains from directly listing shares on a major U.S. stock exchange.

In the case of Mavesa, market efficiency and liquidity was unaffected by NYSE listing. New information was instantaneously and without bias incorporated into Mavesa share prices both before and after NYSE listing, which suggests that there were few efficiency and liquidity gains from direct listing on NYSE.

Although the analysis presented in this paper suggests that the market for Telebras and Astra shares became marginally more efficient following direct listing on NYSE it is difficult to argue that these efficiency and liquidity gains constitute a prima facie argument in favor of direct listing. This paper offers no convincing evidence to suggest that market efficiency or liquidity was significantly improved as a result of directly listing on NYSE as opposed to continued trading in the form of ADRs. On the other hand, it would be difficult to argue that directly listing shares on NYSE did not enhance these companies' visibility, which would greatly facilitate their access to U.S. investment capital.

Finally, the analysis presented in this paper also supports the enormous large body of empirical research that indicates that large, organized U.S. stock exchanges, such as NYSE and AMEX, are weakly efficient and that knowledge of past price movements cannot be used to generate above average returns (See, for example, Moore, 1964; Fama, 1965; Granger & Morgenstern 1963). Although the evidence indicates that trading Telebras and Astra ADRs, which were traded on NYSE, may have been marginally less efficient than trading shares directly listed on NYSE, the rapidity with which new information was incorporated into share prices suggests that it would have been difficult, if not impossible, for even the most savvy and well-informed traders to exploit past price irregularities to generate above normal rates of return.

Suggestions for Further Research

The empirical tests of weak-level market efficiency presented in this paper were generated using daily closing ADR and listed share prices for Telebras, Astra, and Mavesa. An obvious line of further research would be to expand the above analysis to include other foreign companies that initially traded as ADRs but were later directly listed on a major U.S. exchange to determined Whether the above results are representative of a broader pattern in weak-level market efficiency between trading ADRs and directly listed shares. Another suggestion for further research would be to apply the Dickey-Fuller unit root-test for weak-level market efficiency to price data with greater frequency, such as hourly or "tick" data. This is important since the above results may simply be an artifact of the frequency of the data used rather than a reflection of the differences in weak-level efficiency between traded ADRs and foreign company shares traded directly on U.S. stock exchanges.

Endnotes

1. Rule 144-A, Regulation S allows for private placements of sponsored ADRs to qualified institutional buyers (QIBs) to raise capital. Through private placements, foreign companies can raise capital by placing ADRs with large institutional investors, thereby avoiding SEC registration. 144A ADRs trade on Portal in the U.S. No reporting or registration is required and, as such, investors have access to less information about the underlying security. Thus, 144A ADRs are less liquid than registered issues. Incidentally, a Level I ADR program may be established alongside a 144A program (GEIDRS, 1997).

Rule 144-A was adopted by the SEC in 1990 to make U.S. private markets for unregistered securities competitive with European equity markets. By liberalizing re-sales of privately placed securities among QIBs, Rule 144-A enhanced liquidity and lowered the cost of U.S. private placements. A 1993 SEC guideline allowed Regulation S ADRs - those restricted to issue outside of the U.S. - to be traded within the U.S. after a 40-day period of"seasoning."

2. The Wald F-statistic ([F.sub.w])is given by the expression

[F.sub.c] = {(ES[S.sub.R] - ES[S.sub.U])/(m - k)}/{(ES[S.sub.U]/(n-k)}

where ES[S.sub.R] and ES[S.sub.U] represent the error sum of squares for the restricted and unrestricted equations, respectively, n the number of observations, k the number of estimated parameters in the unrestricted equation and m the number of estimated parameters in the restricted equation. The null hypothesis that ? = 0 and ? = 1 is rejected for [F.sub.c] [greater than] [F.sup.*], where [F.sup.*] is the Dickey-Fuller critical F-value. It should be noted that the power of the Dickey-Fuller test is limited in that it only allows us to reject (or fail to reject) the hypothesis that a variable is not a random walk. Failure to reject (especially at a high significance level) is only weak evidence of the random walk hypothesis.

3. Although the Dickey-Fuller test is widely used its power is limited. It only allows us to reject (or fail to reject) the hypothesis that a variable is not a random walk. A failure to reject (especially at a high significance level) is only weak evidence in favor of a random walk hypothesis.

4. The relationship between the dependent and independent variables may under a structural change over time. The Chow test, which is a popular test of structural stability, is given by the expression

[F.sub.c] = {([S.sub.i] - [S.sub.i]')/k}[{[S.sub.i]'/([n.sub.1] + [n.sub.2] - 2k)}.sup.-1]

at k and [n.sub.1]+ [n.sub.2]-2k degrees of freedom, where [n.sub.1] is the number of observations for the first sampling period, [n.sub.2] is the number of observations for the second sampling, k is the number of parameter estimates in the restricted model, [S.sub.i] is the error variations for the entire sampling period, and [S.sub.i]' is the combined error variation for the two sub-periods. This statistic has an F distribution with k and n-2k degrees of freedom (see, for example, Intriligator, Bodkin and Hsiao 1996).

5. Convergence is defined as ([P.sub.t] - [A.sub.1])/[A.sub.2] [not a subset] .001. In other words, share price are said to have converged if the movement of [P.sub.t] to within .001[A.sub.2], of [A.sub.1] following an exogenous shock.

References

Baumol, W. J. (1970). Economic Dynamics. 3rd ed. New York: Macmillan.

Cooper, P. P. (1979). Capital missions: The depositary receipts boom. Global Investment, 9, 18-28.

Dickey, D. A., & Fuller, W. A. (1970). Distribution of the estimators for autoregressive time series with a unit-root. Journal of the American Statistical Association, 74, 427-431.

Dickey, D. A., & Fuller, W. A. (1981). Likelihood ratio statistics for autoregressive time series with a unit-root. Econometrica, 49, 1057-1072.

Fuller, W. A. (1976). Introduction to Statistical Time Series. New York: John Wiley & Sons.

Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work. Journal of Finance, 25, 383-417.

Fama, E. F. (1965). The behavior of stock market prices. Journal of Business, 38, 1, 34-105.

Fama, E. F. (1976). Foundations of finance: Portfolio decisions and securities prices. New York: Basic Books.

Gandolfo, G. (1971). Mathematical methods and models in economic dynamics. London: North-Holland.

Global Equity Investing Depository Receipt Services (GEIDRS). (1997). An overview of depository receipts. The Bank of New York Company, Inc.

Granger, C. W. J., & Morgenstern, O. (1963). Spectral analysis of New York stock market prices. Kyklos, 16, 1-27.

Intriligator, M.D., Bodkin, R.G., & Hsiao, C. (1996). Econometric models, tech

Intriligator, M.D., Bodkin, R.G., & Hsiao, C. (1996) Econometric models, techniques, and applications. Upper Saddle River, NJ: Prentice Hall.

Moore, A.B. (1964) Some characteristics of change in common stock prices. The random character of stock market prices. In Paul H. Cootner, (Ed.), Cambridge, Mass.: M.I.T., pp. 139-161.

Nelson, C.R., & Plosser, C. I. (1982). Trends and random walks in macroeconomic time series: Some evidence and implications. Journal of Monetary Economics, 10, 139-162.