Multi-market trading and arbitrage

Abstract

We measure arbitrage opportunities by comparing the intraday prices and quotes of American Depositary Receipts (ADRs) and other types of cross-listed shares in U.S. markets with synchronous prices of their home-market shares on a currency-adjusted basis for a sample of 506 U.S. cross-listed stocks from 35 different countries. Deviations from price parity average an economically small 4.9 basis points, but they are volatile and can reach large extremes. Price parity deviations and their daily changes are positively related to proxies for holding costs that can impede arbitrage, even after controlling for transactions costs and foreign investment restrictions.

Introduction

Arbitrage is one of the central tenets of financial economics, enforcing the law of one price and keeping markets efficient. It is defined as the “simultaneous purchase and sale of equivalent securities in two different markets in order to profit from discrepancies in their price relationship” (Bodie, Kane, and Marcus, 2009, p. 319). Theoretically, it requires no capital and is without risk. But, in reality, there are important impediments due to market frictions and imperfect information that can limit arbitrage. The market frictions are often explicit and observable, such as transactions costs, taxes, holding costs, and short-sale restrictions. Holding costs are incurred every period that an arbitrage position remains open and include the opportunity cost of capital, the opportunity cost of not receiving full interest on short-sales proceeds, and idiosyncratic risk exposure (Pontiff, 2006). Impediments due to information problems are implicit and more difficult to observe. Another important issue stems from potential agency problems. Many arbitrageurs specialize in this activity and this limits the degree of diversification in their portfolios. If there is a random shock that causes further divergence in an arbitrage position, the specialized arbitrageur cannot diversify away this risk and this limits how much capital he or she invests (Shleifer and Vishny, 1997, De Long et al., 1990). In models without imperfect information or capital constraints (Grossman and Miller, 1988), arbitrageurs are generally more aggressive when prices on the same security diverge further.

Existing research has identified many possible impediments to arbitrage, but, as surveys by Barberis and Thaler (2003) and Lamont and Thaler (2003a) point out, it has done so mostly in the context of a series of “special cases.”¹ These special cases include most commonly studies of closed-end country funds, exchange-traded funds, dual-listed companies (DLCs, or “Siamese twins”), index inclusions/exclusions, and internet carve-outs and spin-offs. While interesting to study, one problem with mispricing phenomena associated with these special cases is that they are, by definition, special. That is, they are limited in scope and number and not always available over extended periods of time. Moreover, in each of these cases, fundamental risk may still exist because arbitrageurs cannot fully hedge their positions due to a lack of perfect substitutes or because the arbitrageur's model may not coincide with the true return-generating process. Our primary contribution is to remedy these two deficiencies in the literature on the limits to arbitrage by investigating the deviations between the prices of shares of stocks that trade simultaneously in different markets around the world. Specifically, we compare the intraday prices and quotes of American depositary receipts (ADRs) and other types of cross-listed shares in U.S. markets with synchronous prices of their home-market shares on a currency-adjusted basis. These stock pairs that trade simultaneously in different markets represent an ideal setting because the shares trade on the same underlying stock and because there are actual market mechanisms in place, such as the conversion/cancellation facilities at depositary banks, for arbitrageurs to be able to exploit these deviations.

Our goal is to determine whether the magnitude of the deviations from price parity for cross-listed pairs varies over time and across different securities in a way that is related to these explicit and implicit arbitrage costs. The sample consists of 506 pairs of cross-listed/home-market shares of stocks from 35 countries over the period between 1993 and 2004. We show that daily deviations from price parity among cross-listed pairs are, on average, economically small averaging around 4.9 basis points (as a percent of the home share price) and with a daily standard deviation of 1.4% for a typical stock pair. But we also find that the level of these deviations from price parity and their daily changes can reach extremes across stock pairs and across time; U.S. share prices relative to home prices range from a discount of as low as −40.4% to a premium as high as 127.4% and the daily changes in the price deviations can decline as much as −95.6% and rise to as high as 168.5%. We also find that these deviations are positively related to holding costs that can impede arbitrage, even after controlling for a variety of transactions costs, taxes, and foreign investment restrictions. The holding cost proxies we consider include the stock pair's idiosyncratic risk that is unrelated to the risk of other securities in either of the competing markets, the stock's dividend yield, which can lower holding costs, and interest rates, which represent an opportunity cost of capital, since arbitrageurs usually do not receive full interest on short-sale proceeds (Pontiff, 1996). Among these proxies, we find that idiosyncratic risk has a statistically reliable, positive relationship to price parity deviations across stock pairs and over time, and it is economically the most important holding cost. Together, these factors explain over 20% of the cross-sectional variation in the levels of and changes in price deviations across the cross-listed pairs.

The market for internationally cross-listed stocks provides a rich but, at the same time, complex setting within which to evaluate the sources of impediments to arbitrage. Consider, for example, the wide variety of instruments for cross-listing shares around the world. ADRs, which are receipts that represent claims against the home-market shares, are the most popular form of cross-listing in the U.S., but there are other types of cross-listed shares (e.g., Global or New York registered shares) which represent direct claims against the same set of risky cash flows as ordinary shares in the home market.² An important distinguishing characteristic of the ADR market is that there exist institutions – namely, depositary banks like the Bank of New York, Citigroup, and JP Morgan Chase – that ensure security “fungibility” by facilitating the convertibility of ADRs into home-market shares and back again. Another special dimension of this experiment is the large cross-section of firms and countries from around the world that have shares cross-listed on U.S. markets. Not only does this allow us as researchers to investigate potential limitations on arbitrage activity at the country level, such as regulatory restrictions, currency controls, foreign ownership limits, and overall equity market development, but it also allows us to consider firm-specific factors that might capture the scope of informational barriers that impede arbitrage activity. Examples include the number of analysts following the stocks, the dispersion in the analysts’ 1-year-ahead earnings forecasts, the fraction of shares held by U.S. institutional investors, and the U.S. market's fraction of the combined shares traded in the competing markets. We find that these proxies for information-based barriers on cross-market arbitrage in internationally cross-listed stocks can explain incrementally another 8–10% of the cross-sectional and time-series variation in the deviations from price parity.

Yet another motivation for our effort is the paucity of papers on arbitrage in the market for ADRs and other types of cross-listed stocks.³ Early studies by Maldonado and Saunders (1983), Kato, Linn, and Schallheim (1991), Wahab, Lashgari, and Cohn (1992), Park and Tavokkol (1994), and Suarez (2005) were limited by a small sample of ADRs from only select countries, like Australia, France, Japan, and the UK, and usually with only weekly prices. Miller and Morey (1996) studied intraday prices for one stock, Glaxo-Wellcome, PLC, during the 2-hour overlap period for their New York Stock Exchange (NYSE) ADRs and UK shares during a 2-month period in 1995. Puthenpurackal (2006) provides a clinical study of Infosys Technologies’ (India) ADR premium around its U.S. offering in 2003.⁴ Rosenthal and Young (1990), Froot and Dabora (1999), Bedi, Richards, and Tennant (2003), and recently De Jong, Rosenthal, and van Dijk (2009) study the special case of dual-listed companies (DLCs, or “Siamese twins”), which effectively represent mergers between companies that agree to combine their operations and cash flows and have common dividend structures while retaining separate shareholder registries and identities, e.g., Royal Dutch and Shell, Unilever N.V. and Unilever PLC. Each of these studies finds large and systematic price parity deviations from their home-market shares which they try to explain with tax, accounting, regulatory, governance, and trading attributes. Recent papers employ special intraday data for country-specific studies of relative price discovery in cross-listed and home-market shares, such as Grammig, Melvin, and Schlag (2005) for three stocks in Germany and Eun and Sabberwal (2003) for 62 stocks in Canada. Finally, Grossmann, Ozuna, and Simpson (2007) examine ADR mispricing for a sample of 74 European firms. Few, if any, of these studies investigate the role of holding cost proxies, like idiosyncratic risk, and none have evaluated how much information-based barriers matter for price deviations in cross-listed stocks.

The next section of the paper outlines some of the important institutional aspects of arbitrage in the market for ADRs and other cross-listed stocks. Our data sample and summary statistics are presented in Section 3. We outline our key hypotheses on the limits to arbitrage in this market, describe the empirical methodology used to test them, and provide the main results in Section 4. Alternative hypotheses are explored in Section 5. A summary of our findings and a discussion about the limitations of our analysis concludes the paper.