Government Intervention and Arbitrage
Pasquariello, Paolo
2017-07-11 00:00:00
Direct government intervention in a market may induce violations of the law of one price in other, arbitrage-related markets. I show that a government pursuing a nonpublic, partially informative price target in a model of strategic market-order trading and segmented dealership generates equilibrium price differentials among fundamentally identical assets by clouding dealers’ inference about the targeted asset’s payoff from its order flow, to an extent complexly dependent on existing price formation. I find supportive evidence using a sample of American Depositary Receipts and other cross-listings traded in the major U.S. exchanges, along with currency interventions by developed and emerging countries between 1980 and 2009. Received May 18, 2016; editorial decision May 10, 2017 by Editor Andrew Karolyi. Modern finance rests on the law of one price (LOP). The LOP states that unimpeded arbitrage activity should eliminate price differences for identical assets in well-functioning markets. The study of frictions leading to LOP violations is crucial to understanding the forces affecting the quality of the process of price formation in financial markets—their ability to price assets correctly on an absolute and relative basis. Accordingly, the literature reports evidence of LOP violations in several financial markets, often explains their occurrence and intensity with unspecified behavioral or (less often and anecdotally) rational demand shocks unrelated to asset fundamentals, and attributes their persistence to various limits to arbitrageurs’ efforts to fully absorb those shocks (e.g., Shleifer 2000; Lamont and Thaler 2003; Gromb and Vayanos 2010). I contribute to this understanding by investigating the role of a specific and empirically observable form of rational demand shocks—direct government intervention—in the emergence of LOP violations, ceteris paribus for limits to arbitrage. Central banks and governmental agencies (“governments” for brevity) routinely trade securities in pursuit of economic and financial policy.1 Recently, both the scale and frequency of this activity have soared in the aftermath of the financial crisis of 2008–2009. The pursuit of policy via “official” trading in financial assets has long been found both to be effective and to yield welfare gains, for example, by achieving “intermediate” monetary targets (Rogoff 1985; Corrigan and Davis 1990; Edison 1993; Sarno and Taylor 2001; Hassan, Mertens, and Zhang 2016). I model and document the novel notion that such government intervention may also induce LOP violations and so worsen financial market quality. My analysis indicates that these price distortions in the affected markets may be nontrivial and hence may have nontrivial effects on their allocational and risk-sharing roles. The insight that direct government intervention in financial markets can create negative externalities on their quality has important implications for the broader debate on financial stability, optimal financial regulation, and unconventional policy making (e.g., Acharya and Richardson 2009; Hanson, Kashyap, and Stein 2011; Bernanke 2012).2 I illustrate this notion within a standard, parsimonious one-period model of strategic multiasset trading based on Kyle (1985) and Chowdhry and Nanda (1991). In the economy’s basic setting, two fundamentally identical, or linearly related risky assets—labeled $$1$$ and $$2$$—are exchanged by three types of risk-neutral market participants: a discrete number of heterogeneously informed multiasset speculators, single-asset noise traders, and competitive market makers. If the dealership sector is segmented, market makers in one asset do not observe order flow in the other asset (e.g., Subrahmanyam 1991a; Baruch, Karolyi, and Lemmon 2007; Boulatov, Hendershott, and Livdan 2013). Then liquidity demand differentials from less-than-perfectly correlated noise trading in assets $$1$$ and $$2$$ yield equilibrium LOP violations (i.e., less-than-perfectly correlated equilibrium prices of these assets) despite semi-strong efficiency in either market and fundamentally informed, hence perfectly correlated speculation across both (e.g., like in Chowdhry and Nanda 1991). Intuitively, those relative mispricings—nonzero price differentials—can occur in equilibrium because speculators can only submit camouflaged market orders in each asset, that is, together with noise traders and before market-clearing prices are set. Accordingly, when both markets are more illiquid, noise trading in either asset has a greater impact on its equilibrium price, yielding larger LOP violations. Dealership segmentation, speculative market-order trading, and liquidity demand differentials in the model serve as a reduced-form representation of existing forces behind LOP violations and impediments to arbitrage activity in financial markets. In this setting, I introduce a stylized government submitting camouflaged market orders (e.g., Vitale 1999; Naranjo and Nimalendran 2000) in only one of the two assets, asset $$1$$, in pursuit of policy—a nonpublic, partially informative price target (e.g., Bhattacharya and Weller 1997). I then show that such government intervention increases equilibrium LOP violations, that is, lowers the equilibrium price correlation of assets $$1$$ and $$2$$, ceteris paribus for those limits to arbitrage and even in the absence of liquidity demand differentials. An intuitive explanation for this result is that the uncertainty surrounding the government’s intervention policy in asset $$1$$ clouds the inference of the market makers about its fundamentals when setting the equilibrium price of that asset from its order flow. Consistently, the magnitude of this effect is increasing in government policy uncertainty and generally, yet not uniformly decreasing in pre-intervention market quality. In particular, intervention-induced LOP violations are larger when market liquidity is low, for example, in the presence of more heterogeneously informed speculators or less intense noise trading, since in those circumstances official trading has a greater impact on the equilibrium price of asset $$1$$. However, intervention-induced LOP violations may also be complexly related to extant such violations. For example, they may be larger in the presence of fewer speculators yet smaller in the presence of less correlated noise trading, since in the former circumstances official trading has a greater impact on the already low equilibrium price correlation of assets $$1$$ and $$2$$ than in the latter. I test the model’s main implications by examining the impact of government interventions in the foreign exchange (“forex”) market on LOP violations in the U.S. market for American Depositary Receipts and other cross-listed stocks (“ADRs” for brevity). The forex market is one of the largest, most liquid financial markets in the world (e.g., Bank for International Settlements 2016). The major U.S. exchanges (the “ADR market”) are the most important venue for international cross-listings (e.g., Karolyi 1998, 2006). These markets also serve as a setting that is as close as possible in spirit to the assumptions in my model. First, an ADR is a dollar-denominated security, traded in the United States, representing a set number of shares in a foreign stock held in deposit by a U.S. financial institution; hence, its price is linked to the underlying exchange rate by an arbitrage relation, the “ADR parity” (ADRP; e.g., Gagnon and Karolyi 2010; Pasquariello, Roush, and Vega 2014). This fundamental linkage can be described in my setting as a linear relation between the terminal payoff of asset $$1$$, the exchange rate (traded in the forex market), and the terminal payoff of asset $$2$$, the ADR (traded in the U.S. stock market). My model then predicts that, ceteris paribus, forex intervention (government intervention targeting the price of asset 1) may induce ADRP violations, that is, lowers the equilibrium correlation between the price of the actual ADR (asset $$2$$) and its synthetic, arbitrage-free price implied by the ADRP (a linear function of the price of asset $$1$$). Second, forex and ADR dealership sectors are arguably less-than-perfectly integrated, as market makers in either market are less likely to observe order flow in the other market. Third, according to the literature (surveyed in Edison 1993; Sarno and Taylor 2001; Neely 2005; Menkhoff 2010; Engel 2014), government intervention in currency markets is common and often secret; its policy objectives are often nonpublic; its effectiveness is statistically robust and often attributed to their perceived informativeness about fundamentals. Fourth, most forex interventions are sterilized (i.e., do not affect the money supply of the targeted currencies), and all of them are unlikely to be prompted by ADRP violations. I construct a sample that includes ADRs traded in the major U.S. exchanges as well as official trading activity of developed and emerging countries in the currency markets between 1980 and 2009. Its salient features are in line with the aforementioned literature. Average absolute percentage ADRP violations are large (e.g., a $$2\%$$ [$$200$$ basis points, bps] deviation from the arbitrage-free price), generally decline as financial integration increases, but display meaningful intertemporal dynamics (e.g., spiking during periods of financial instability). Forex interventions are also nontrivial, albeit small relative to average turnover in the currency markets, are especially frequent between the mid-1980s and the mid-1990s, and typically involve exchange rates relative to the dollar. The empirical analysis of this sample provides support for my model. I find that measures of the actual and historically abnormal intensity of ADRP violations increase in measures of the actual and historically abnormal intensity of forex interventions. This relation is both statistically and (plausibly) economically significant. For instance, a one-standard-deviation increase in forex intervention activity in a month is accompanied by a material average cumulative increase in absolute ADRP violations of up to $$10$$ bps, which is as much as $$45\%$$ of the sample volatility of their monthly changes. This relation is also robust to controlling for several proxies for market conditions that are commonly associated with LOP violations, limits to arbitrage, and/or forex intervention (e.g., Pontiff 1996, 2006; Pasquariello 2008, 2014; Gagnon and Karolyi 2010; Garleanu and Pedersen 2011; Engel 2014), as well as to removing ADRs from emerging countries from the analysis when affected by the imposition of capital controls (e.g., Edison and Warnock 2003; Auguste et al. 2006). Importantly, those same official currency trades are not accompanied by larger LOP violations in the much more closely integrated currency and international money markets in many respects, including dealership (e.g., McKinnon 1977; Dufey and Giddy 1994; Bekaert and Hodrick 2012), as they are unrelated to violations of the covered interest rate parity (CIRP), an arbitrage relation between interest rates and spot and forward exchange rates commonly used to proxy for currency market quality (e.g., Frenkel and Levich 1975, 1977; Coffey et al. 2009; Griffoli and Ranaldo 2011). This finding not only is consistent with my model but also suggests that my results are unlikely to stem from a dislocation in currency markets leading to both forex interventions and ADRP violations (e.g., Neely and Weller 2007). Further cross-sectional and time-series analysis indicates that poor, deteriorating price formation in the ADR arbitrage-linked markets magnify ADRP violations both directly and through its possibly complex linkage with forex intervention activity, as postulated by my model. In particular, I find LOP violations to be larger and the linkage to be stronger not only for ADRs from emerging economies but also for markets and portfolios of ADRs of high underlying quality, as well as in correspondence with high or greater ADRP illiquidity (as measured by the average fraction of zero returns in the currency, U.S., and foreign stock markets), greater dispersion of beliefs about common fundamentals (as measured by the standard deviation of professional forecasts of U.S. macroeconomic news releases), and greater uncertainty about governments’ currency policy (as measured by real-time intervention volatility). For example, the positive estimated impact of high forex intervention activity on ADRP violations is more than three times larger when in correspondence with high information heterogeneity among market participants. In summary, my study highlights novel, and potentially important, adverse implications of direct government intervention, a frequently employed instrument of policy with well-understood benefits, for financial market quality. 1. Theory I am interested in the effects of government intervention on relative mispricings, that is, on LOP violations. To that purpose, I first describe, in Section 1.1, a standard noisy rational expectations equilibrium (REE) model of multiasset informed trading. The model, based on Kyle (1985), is a straightforward extension of Chowdhry and Nanda (1991) to imperfectly competitive speculation and nondiscretionary liquidity trading that allows for relative mispricings in equilibrium. I then contribute to the literature on limits to arbitrage, in Section 1.2, by introducing in this setting a stylized government and considering the implications of its official trading activity for LOP violations. The Appendix contains all proofs. 1.1 The basic model of multiasset trading The basic model is based on Kyle (1985) and Chowdhry and Nanda (1991). The model’s standard framework has often been used to study price formation in many financial markets and for many asset classes (see, e.g., the surveys in O’Hara 1995; Vives 2008; Foucault, Pagano, and Röell 2013). It is a two-date ($$t=0,1$$) economy in which two risky assets ($$i=1,2$$) are exchanged. Trading occurs only at date $$t=1$$, after which each asset’s payoff $$v_{i}$$ is realized. The two assets are fundamentally related in that $$v_{i}\equiv a_{i}+b_{i}v$$, where $$v$$ is normally distributed with mean $$p_{0}$$ and variance $$\sigma _{v}^{2}$$, and $$a_{i}$$ and $$b_{i}$$ are constants. Fundamental commonality in payoffs is meant to parsimoniously represent a wide range of LOP relations between the two assets; linearity of their payoffs in $$v$$ ensures that the model can be solved in closed form. I discuss one particular such representation for the ADR parity in Section 2.1. For simplicity and without loss of generality, I assume that the two assets are fundamentally identical in that $$a_{i}=0$$ and $$b_{i}=1$$, such that $$v_{i}=v$$. There are three types of risk-neutral traders: a discrete number ($$M$$) of informed traders (labeled speculators) in both assets (e.g., Foucault and Gehrig 2008; Pasquariello and Vega 2009), as well as nondiscretionary liquidity traders and competitive market makers (MMs) in each asset. All traders know the structure of the economy and the decision-making process leading to order flow and prices. At date $$t=0$$, there is neither information asymmetry about $$v$$ nor trading. Sometime between $$t=0$$ and $$t=1$$, each speculator $$m$$ receives a private and noisy signal of $$v$$, $$S_{v}\left(m\right) $$. I assume that each signal $$S_{v}\left( m\right) $$ is drawn from a normal distribution with mean $$p_{0}$$ and variance $$\sigma _{s}^{2}$$ and that, for any two $$S_{v}\left( m\right) $$ and $$S_{v}\left( j\right) $$, $$cov\left[ v,S_{v}\left( m\right) \right] =cov\left[ S_{v}\left( m\right),S_{v}\left( j\right) \right] =\sigma _{v}^{2}$$. Each speculator’s information endowment about $$v$$ is defined as $$\delta _{v}\left( m\right) \equiv E\left[ v|S_{v}\left( m\right) \right] -p_{0}$$. I characterize speculators’ private information heterogeneity by further imposing that $$\sigma _{s}^{2}=\frac{1}{\rho }\sigma _{v}^{2}$$ and $$\rho \in \left( 0,1\right) $$. This parsimonious parametrization implies that $$\delta _{v}\left( m\right) =\rho \left[ S_{v}\left( m\right) -p_{0}\right] $$ and $$E\left[ \delta _{v}\left( j\right) |\delta _{v}\left( m\right) \right] =\rho \delta _{v}\left( m\right) $$, i.e., that $$\rho $$ is the unconditional correlation between any two $$\delta _{v}\left( m\right) $$ and $$\delta _{v}\left( j\right) $$. Intuitively, as $$\rho $$ declines, speculators’ private information about $$v$$ becomes more dispersed, thus is less precise and correlated.3 At date $$t=1$$, speculators and liquidity traders submit their orders in assets $$1$$ and $$2$$ to the MMs before their equilibrium prices $$p_{1,1}$$ and $$p_{1,2}$$ have been set. The market order of each speculator $$m$$ in each asset $$i$$ is defined as $$x_{i}\left( m\right) $$, such that her profit is given by $$\pi \left( m\right) =\left( v-p_{1,1}\right) x_{1}\left( m\right) +\left( v-p_{1,2}\right) x_{2}\left( m\right) $$. Liquidity traders generate random, normally distributed demands $$z_{1}$$ and $$z_{2}$$, with mean zero, variance $$\sigma _{z}^{2}$$, and covariance $$\sigma _{zz}$$, where $$\sigma _{zz}\in \left( 0,\sigma _{z}^{2}\right] $$.4 For simplicity, $$z_{1}$$ and $$z_{2}$$ are assumed to be independent from all other random variables. Competitive MMs in each asset $$i$$ do not receive any information about its terminal payoff $$v$$, and observe only that asset’s aggregate order flow, $$\omega _{i}=\sum_{m=1}^{M}x_{i}\left( m\right) +z_{i}$$, before setting the market-clearing price, $$p_{1,i}=p_{1,i}\left( \omega _{i}\right) $$, like in Chowdhry and Nanda (1991), Subrahmanyam (1991a), Baruch, Karolyi, and Lemmon (2007), Pasquariello and Vega (2009), and Boulatov, Hendershott, and Livdan (2013). Segmentation in market making is an important feature of the model, as it allows for the possibility that $$ p_{1,1}$$ and $$p_{1,2}$$ might be different in equilibrium despite assets $$1$$ and $$2$$’s identical payoffs.5 1.1.1 Equilibrium A Bayesian Nash equilibrium of this economy is a set of $$2\left( M+1\right) $$ functions, $$x_{i}\left( m\right) \left( \cdot \right) $$ and $$p_{1,i}\left( \cdot \right) $$, satisfying the following conditions: Utility maximization: $$x_{i}\left( m\right) \left( \delta _{v}\left( m\right) \right) =\arg \max E\left[ \pi \left( m\right) |\delta _{v}\left( m\right) \right] $$; Semi-strong market efficiency: $$p_{1,i}=E\left( v|\omega _{i}\right) $$.6 Proposition 1 describes the unique linear REE that obtains. Proposition 1. There exists a unique linear equilibrium given by the price functions: \begin{equation} p_{1,i}=p_{0}+\lambda \omega _{i}\text{,} \label{price} \end{equation} (1) where $$\lambda =\frac{\sigma _{v}\sqrt{M\rho }}{\sigma _{z}\left[ 2+\left( M-1\right) \rho \right] }>0$$; and by each speculator’s orders: \begin{equation} x_{i}\left( m\right) =\frac{\sigma _{z}}{\sigma _{v}\sqrt{M\rho }}\delta _{v}\left( m\right) \text{.} \label{trade} \end{equation} (2) In this class of models, MMs in each market $$i$$ learn about the traded asset $$i$$’s terminal payoff from its order flow, $$\omega _{i}$$; hence, each imperfectly competitive, risk-neutral speculator trades cautiously in both assets ($$\left\vert x_{i}\left( m\right) \right\vert <\infty $$, Equation (2)) to protect the information advantage stemming from her private signal, $$S_{v}\left(m\right) $$. Like in Kyle (1985), positive equilibrium price impact or lambda ($$\lambda >0$$) compensates the MMs for their expected losses from speculative trading in $$\omega _{i}$$ with expected profits from noise trading ($$z_{i}$$). The ensuing comparative statics are intuitive and standard in the literature (e.g., Subrahmanyam 1991b; Pasquariello and Vega 2009). MMs’ adverse selection risk is more severe and equilibrium liquidity lower in both markets (higher $$\lambda $$) when: (1) the traded assets’ identical terminal payoff $$v$$ is more uncertain (higher $$\sigma _{v}^{2}$$), since speculators’ private information advantage is greater; (2) their private signals are less correlated (lower $$\rho $$), since each of them, perceiving to have greater monopoly power on her private information, trades more cautiously with it (lower $$\left\vert x_{i}\left( m\right) \right\vert $$); (3) noise trading is less intense (lower $$\sigma _{z}^{2}$$), since MMs need to be compensated for less camouflaged speculation in the order flow; or (4) there are fewer speculators in the economy (lower $$M$$), since imperfect competition among them magnifies their cautious aggregate trading behavior (lower $$\left\vert \sum_{m=1}^{M}x_{i}\left( m\right) \right\vert $$).7 1.1.2 LOP violations The literature defines and measures LOP violations either as nonzero price differentials or as less-than-perfect price correlations among identical assets (e.g., Karolyi 1998, 2006; Auguste et al. 2006; Pasquariello 2008, 2014; Gagnon and Karolyi 2010; Gromb and Vayanos 2010; Griffoli and Ranaldo 2011). As I further discuss in Section 2.1.1, the two representations are conceptually equivalent in the economy. An examination of Equations (1) and (2) in Proposition 1 reveals that less-than-perfectly correlated noise trading in assets $$1$$ and $$2$$ ($$\sigma _{zz}<\sigma _{z}^{2}$$) may lead to nonzero realizations of liquidity demand ($$z_{1}\neq z_{2}$$) and price differentials ($$p_{1,1}\neq p_{1,2}$$) in equilibrium, by at least partly offsetting fundamentally informed (i.e., perfectly correlated) trading in those assets ($$x_{1}\left( m\right) =x_{2}\left( m\right) $$). Of course, this may occur only with segmented market making allowing for $$E\left( v|\omega _{1}\right) \neq E\left( v|\omega _{2}\right) $$. If MMs observe order flow in both assets (i.e., with perfectly integrated market making), no price differential can arise in equilibrium since semi-strong market efficiency in Condition 2 implies that $$p_{1,1}=E\left( v|\omega _{1},\omega _{2}\right) =p_{1,2}$$. I formalize these observations in Corollary 1 by measuring LOP violations in the economy using the unconditional correlation of the equilibrium prices of assets $$1$$ and $$2$$, $${\it corr}\left( p_{1,1},p_{1,2}\right) $$, like in Gromb and Vayanos (2010). Corollary 1. In the presence of less-than-perfectly correlated noise trading, the LOP is violated in equilibrium: \begin{equation} {\it corr}\left( p_{1,1},p_{1,2}\right) =1-\frac{\sigma _{z}^{2}-\sigma _{zz}}{ \sigma _{z}^{2}\left[ 2+\left( M-1\right) \rho \right] }<1\text{.} \label{correlation} \end{equation} (3) There are no LOP violations under perfectly integrated market making or perfectly correlated noise trading. Figures 1 and 2 illustrate the intuition behind Corollary 1. I consider a baseline economy in which $$\sigma _{v}^{2}=1$$, $$\sigma _{z}^{2}=1$$, $$\sigma _{zz}=0.5$$, $$\rho =0.5$$, and $$M=10$$. I then plot the equilibrium price correlation of Equation (3) as a function of $$\sigma _{zz}$$, $$\rho $$, $$M$$, or $$\sigma _{z}^{2}$$ in Figures 1A to 1D, respectively (solid lines). Figure 2A displays (average) $${\it corr}\left( p_{1,1},p_{1,2}\right) $$ as a function of the corresponding (average) $$\lambda $$ for both the relation between $${\it corr}\left( p_{1,1},p_{1,2}\right) $$ and $$\rho $$ of Figure 1B (solid line, right axis, for $$\sigma _{z}^{2}=1$$ and $$\rho \approx 0.5$$) and the relation between $${\it corr}\left( p_{1,1},p_{1,2}\right) $$ and $$\sigma _{z}^{2}$$ of Figure 1D (dashed line, left axis, for $$\rho =0.5$$ and $$\sigma _{z}^{2}\approx 1$$).8 Figure 1 View largeDownload slide LOP violations and model parameters This figure plots the unconditional correlation between the equilibrium prices of assets $$1$$ and $$2$$ in the absence ($${\it corr}( p_{1,1},p_{1,2})$$ of Equation (3), solid lines) and in the presence of government intervention ($${\it corr}( p_{1,1}^{\ast},p_{1,2}^{\ast })$$ of Equation (10), dashed lines), as a function of either $$\sigma _{zz}$$ (the covariance of noise trading in those assets, in Figure 1A), $$\rho $$ (the correlation of speculators’ private signals $$S_{v}(m) $$ about $$v$$, the identical terminal payoff of assets $$1$$ and $$2$$, in Figure 1B), $$M$$ (the number of speculators, in Figure 1C), $$\sigma_{z}^{2}$$ (the intensity of noise trading, in Figure 1D), $$\gamma$$ (the government’s commitment to its policy target $$p_{1,1}^{T}$$ for the equilibrium price of asset $$1$$ in its loss function $$L({\it gov}) $$ of Equation (4)), $$\mu $$ (the correlation of the government’s policy target $$p_{1,1}^{T}$$ with its private signal $$S_{v}({\it gov}) $$ about the identical terminal payoff $$v$$ of assets $$1$$ and $$2$$), $$\psi $$ (the precision of the government’s private signal of $$v$$, $$S_{v}({\it gov}) $$), and $$\sigma _{v}^{2}$$ (the uncertainty about $$v$$, the identical terminal payoff of assets $$1$$ and $$2$$, in Figure 1H), when $$\sigma_{v}^{2}=1$$, $$\sigma _{z}^{2}=1$$, $$\sigma _{zz}=0.5$$, $$\rho =0.5$$, $$\psi =0.5$$, $$\gamma =0.5$$, $$\mu =0.5$$, and $$M=10$$. Figure 1 View largeDownload slide LOP violations and model parameters This figure plots the unconditional correlation between the equilibrium prices of assets $$1$$ and $$2$$ in the absence ($${\it corr}( p_{1,1},p_{1,2})$$ of Equation (3), solid lines) and in the presence of government intervention ($${\it corr}( p_{1,1}^{\ast},p_{1,2}^{\ast })$$ of Equation (10), dashed lines), as a function of either $$\sigma _{zz}$$ (the covariance of noise trading in those assets, in Figure 1A), $$\rho $$ (the correlation of speculators’ private signals $$S_{v}(m) $$ about $$v$$, the identical terminal payoff of assets $$1$$ and $$2$$, in Figure 1B), $$M$$ (the number of speculators, in Figure 1C), $$\sigma_{z}^{2}$$ (the intensity of noise trading, in Figure 1D), $$\gamma$$ (the government’s commitment to its policy target $$p_{1,1}^{T}$$ for the equilibrium price of asset $$1$$ in its loss function $$L({\it gov}) $$ of Equation (4)), $$\mu $$ (the correlation of the government’s policy target $$p_{1,1}^{T}$$ with its private signal $$S_{v}({\it gov}) $$ about the identical terminal payoff $$v$$ of assets $$1$$ and $$2$$), $$\psi $$ (the precision of the government’s private signal of $$v$$, $$S_{v}({\it gov}) $$), and $$\sigma _{v}^{2}$$ (the uncertainty about $$v$$, the identical terminal payoff of assets $$1$$ and $$2$$, in Figure 1H), when $$\sigma_{v}^{2}=1$$, $$\sigma _{z}^{2}=1$$, $$\sigma _{zz}=0.5$$, $$\rho =0.5$$, $$\psi =0.5$$, $$\gamma =0.5$$, $$\mu =0.5$$, and $$M=10$$. Figure 2 View largeDownload slide LOP violations and other equilibrium outcomes This figure plots the average unconditional equilibrium correlation between the equilibrium prices of assets} $$1$$ {and} $$2$$ in the absence ($${\it corr}(p_{1,1},p_{1,2})$$ of Equation (3)) and in the presence of government intervention ($${\it corr}( p_{1,1}^{\ast },p_{1,2}^{\ast })$$ of Equation (10)), $$\overline{{\it corr}( p_{1,1},p_{1,2}) }=\frac{1}{2}[ {\it corr}( p_{1,1},p_{1,2}) +{\it corr}( p_{1,1}^{\ast },p_{1,2}^{\ast }) ] $$, as a function of the corresponding average equilibrium price impact $$\overline{\lambda }=\frac{1}{2}( \lambda +\lambda ^{\ast }) $$ (in Figure 2A) as well as their difference, $$\Delta {\it corr}( p_{1,1},p_{1,2}) \equiv {\it corr}( p_{1,1},p_{1,2}) -{\it corr}( p_{1,1}^{\ast },p_{1,2}^{\ast }) $$, as a function of either the corresponding $$\overline{\lambda }$$ (in Figure 2B) or the corresponding $$ {\it corr}( p_{1,1},p_{1,2}) $$ (in Figure 2C), for both the relation between $${\it corr}( p_{1,1},p_{1,2}) $$ and $$\rho $$ of Figure 1B for $$\sigma _{z}^{2}=1$$ and $$\rho \approx 0.5$$ (solid line, right axis) and the relation between $${\it corr}( p_{1,1},p_{1,2}) $$ and $$\sigma _{z}^{2}$$ of Figure 1D for $$\rho =0.5$$ and $$\sigma _{z}^{2}\approx 1$$ (dashed line, left axis), when $$\sigma _{v}^{2}=1$$, $$\sigma _{z}^{2}=1$$, $$\sigma _{zz}=0.5$$, $$\rho =0.5$$, $$\psi =0.5$$, $$\gamma =0.5$$, $$\mu =0.5$$, and $$M=10$$. Figure 2 View largeDownload slide LOP violations and other equilibrium outcomes This figure plots the average unconditional equilibrium correlation between the equilibrium prices of assets} $$1$$ {and} $$2$$ in the absence ($${\it corr}(p_{1,1},p_{1,2})$$ of Equation (3)) and in the presence of government intervention ($${\it corr}( p_{1,1}^{\ast },p_{1,2}^{\ast })$$ of Equation (10)), $$\overline{{\it corr}( p_{1,1},p_{1,2}) }=\frac{1}{2}[ {\it corr}( p_{1,1},p_{1,2}) +{\it corr}( p_{1,1}^{\ast },p_{1,2}^{\ast }) ] $$, as a function of the corresponding average equilibrium price impact $$\overline{\lambda }=\frac{1}{2}( \lambda +\lambda ^{\ast }) $$ (in Figure 2A) as well as their difference, $$\Delta {\it corr}( p_{1,1},p_{1,2}) \equiv {\it corr}( p_{1,1},p_{1,2}) -{\it corr}( p_{1,1}^{\ast },p_{1,2}^{\ast }) $$, as a function of either the corresponding $$\overline{\lambda }$$ (in Figure 2B) or the corresponding $$ {\it corr}( p_{1,1},p_{1,2}) $$ (in Figure 2C), for both the relation between $${\it corr}( p_{1,1},p_{1,2}) $$ and $$\rho $$ of Figure 1B for $$\sigma _{z}^{2}=1$$ and $$\rho \approx 0.5$$ (solid line, right axis) and the relation between $${\it corr}( p_{1,1},p_{1,2}) $$ and $$\sigma _{z}^{2}$$ of Figure 1D for $$\rho =0.5$$ and $$\sigma _{z}^{2}\approx 1$$ (dashed line, left axis), when $$\sigma _{v}^{2}=1$$, $$\sigma _{z}^{2}=1$$, $$\sigma _{zz}=0.5$$, $$\rho =0.5$$, $$\psi =0.5$$, $$\gamma =0.5$$, $$\mu =0.5$$, and $$M=10$$. LOP violations are larger when noise trading in assets $$1$$ and $$2$$ is less correlated (lower $$\sigma _{zz}$$ in Figure 1A), since liquidity demand and price differentials are more likely in equilibrium (e.g., like in Chowdhry and Nanda 1991). LOP violations are also larger when equilibrium liquidity in both markets is lower (i.e., the higher is $$\lambda $$), since the impact of noise trading on equilibrium prices is greater and the price differentials stemming from liquidity demand differentials in Equation (1) are larger. Thus, $${\it corr}\left( p_{1,1},p_{1,2}\right) $$ is greater when there are fewer speculators in the economy (lower $$M$$ in Figure 1B) or when their private information is more dispersed (lower $$\rho $$ in Figures 1C to 2A), since the more cautious is their (aggregate or individual) trading activity and the more serious is the threat of adverse selection for MMs.9 Lastly, more intense noise trading (higher $$\sigma _{z}^{2}$$ in Figures 1D to 2A) amplifies LOP violations by increasing both the likelihood and magnitude of liquidity demand differentials, despite its lesser impact (via lower $$\lambda$$) on equilibrium prices. I summarize these observations in Corollary 2. Corollary 2. LOP violations increase in speculators’ information heterogeneity and the intensity of noise trading, as well as decrease in the number of speculators and the covariance of noise trading. LOP violations do not necessarily imply riskless arbitrage opportunities. While the former occur whenever nonzero price differences between two assets with identical liquidation value arise, the latter require that those differences be exploitable with no risk. In my setting, only speculators can and do trade strategically and simultaneously in both assets $$1$$ and $$2$$ (see Equation (2)). Hence, only they can attempt to profit from any price difference they anticipate to observe. However, the unconditional expected prices of assets $$1$$ and $$2$$ are identical in equilibrium ($$E\left( p_{1,1}\right) =E\left( p_{1,2}\right) $$) since, by Condition 2, both $$p_{1,1}$$ and $$p_{1,2}$$ incorporate all individual private information about their identical terminal value $$v$$ (i.e., all private signals $$S_{v}\left( m\right) $$ in Equation (1)). Further, speculators cannot place limit orders and, in the noisy REE of Proposition 1, neither observe nor can accurately predict the market-clearing prices of assets $$1$$ and $$2$$ when submitting their market orders, $$x_{i}\left( m\right) $$. Thus, there is no feasible riskless arbitrage opportunity in the economy.10 Segmentation in market making, speculative market-order trading, and less-than-perfectly correlated noise trading in the basic model are a reduced-form representation of existing forces affecting the ability of financial markets to correctly price assets that are fundamentally linked by an arbitrage parity. 1.2 Government intervention Governments often intervene in financial markets. The trading activity of central banks and various governmental agencies has been argued and shown both to affect price levels and dynamics of exchange rates, sovereign bonds, derivatives, and stocks, as well as to yield often conflicting microstructure externalities. Recent studies include Bossaerts and Hillion (1991), Dominguez and Frankel (1993), Bhattacharya and Weller (1997), Vitale (1999), Naranjo and Nimalendran (2000), Lyons (2001), Dominguez (2003), 2006), Evans and Lyons (2005), and Pasquariello (2007b, 2010) for the spot and forward currency markets, Harvey and Huang (2002), Ulrich (2010), Brunetti, di Filippo, and Harris (2011), D’Amico and King (2013), Pasquariello, Roush, and Vega (2014), and Pelizzon et al. (2016) for the money and bond markets, and Sojli and Tham (2010) and Dyck and Morse (2011) for the stock markets.11 As such, this “official” trading activity may have an impact on the ability of the affected markets to price assets correctly. I explore this possibility by introducing a stylized government in the multiasset economy of Section 1.1. The literature identifies several recurring features of direct government intervention in financial markets (e.g., Edison 1993; Vitale 1999; Sarno and Taylor 2001; Neely 2005; Menkhoff 2010; Engel 2014; Pasquariello, Roush, and Vega 2014): (1) governments tend to pursue nonpublic price targets in those markets; (2) governments often intervene in secret in the targeted markets; (3) governments are likely or perceived to have an information advantage over most market participants about the fundamentals of the traded assets; (4) the observed ex post effectiveness of governments at pursuing their price targets is often attributed to that actual or perceived information advantage; (5) those price targets may be related to governments’ fundamental information; and (6) governments are sensitive to the potential costs of their interventions. I parsimoniously capture these features using the following assumptions about the stylized government. First, the government is given a private and noisy signal of $$v$$, $$S_{v}\left({\it gov}\right) $$, a normally distributed variable with mean $$p_{0}$$, variance $$\sigma _{gov}^{2}=\frac{1}{\psi }\sigma _{v}^{2}$$, and precision $$\psi \in \left( 0,1\right) $$. I further impose that $$cov\left[ S_{v}\left( m\right), S_{v}\left({\it gov}\right) \right] =cov\left[ v,S_{v}\left({\it gov}\right) \right] =\sigma _{v}^{2}$$, as for speculators’ private signals $$S_{v}\left(m\right) $$ in Section 1.1. Accordingly, the government’s information endowment about $$v$$ is defined as $$\delta _{v}\left({\it gov}\right) \equiv E\left[ v|S_{v}\left({\it gov}\right) \right] -p_{0}=\psi \left[ S_{v}\left({\it gov}\right) -p_{0}\right] $$. Second, the government is given a nonpublic target for the price of asset $$1$$, $$p_{1,1}^{T}$$, drawn from a normal distribution with mean $$\overline{p}_{1,1}^{T}$$ and variance $$\sigma _{T}^{2}$$. The government’s information endowment about $$p_{1,1}^{T}$$ is then $$\delta _{T}\left({\it gov}\right) \equiv p_{1,1}^{T}-\overline{p}_{1,1}^{T}$$.12 This policy target is some unspecified function of $$S_{v}\left({\it gov}\right) $$, such that $$\sigma _{T}^{2}=\frac{1}{\mu }\sigma _{gov}^{2}=\frac{1}{\mu \psi }\sigma _{v}^{2}$$, $${\it cov}\left[ p_{1,1}^{T},S_{v}\left({\it gov}\right) \right] =\sigma _{\it gov}^{2}$$, and $${\it cov}\left[ S_{v}\left( m\right), p_{1,1}^{T}\right] ={\it cov}\left( v,p_{1,1}^{T}\right) =\sigma _{v}^{2}$$. Hence, when $$\mu \in \left( 0,1\right) $$ is higher, the government’s price target is more correlated to its fundamental information and market participants are less uncertain about its policy. For example, this assumption captures the observation that government interventions in currency markets either “chase the trend” (if $$\mu$$ is high) to reinforce market participants’ beliefs about fundamentals as reflected by observed exchange rate dynamics (e.g., Edison 1993; Sarno and Taylor 2001; Engel 2014) or more often “lean against the wind” (if $$\mu$$ is low) to resist those beliefs and dynamics (e.g., Lewis 1995; Kaminsky and Lewis 1996; Bonser-Neal, Roley, and Sellon 1998; Pasquariello 2007b).13 Third, the government can only trade in asset $$1$$; at date $$t=1$$, before the equilibrium price $$p_{1,1}$$ has been set, it submits to the MMs a market order $$x_{1}\left({\it gov}\right) $$ minimizing the expected value of its loss function: \begin{equation} L\left({\it gov}\right) =\gamma \left( p_{1,1}-p_{1,1}^{T}\right) ^{2}+\left( 1-\gamma \right) \left( p_{1,1}-v\right) x_{1}\left({\it gov}\right) \text{,} \label{loss} \end{equation} (4) where $$\gamma \in \left( 0,1\right) $$. This specification is based on Stein (1989), Bhattacharya and Weller (1997), Vitale (1999), and Pasquariello, Roush, and Vega (2014). The first term in Equation (4) is meant to capture the government’s attempts to achieve its policy objectives for asset $$1$$ by trading to minimize the squared distance between asset $$1$$’s equilibrium price, $$p_{1,1}$$, and the target, $$p_{1,1}^{T}$$. The second term in Equation (4) accounts for the costs of that intervention, namely, deviating from pure profit-maximizing speculation in asset $$1$$ ($$\gamma =0$$). When $$\gamma $$ is higher, the government is more committed to policy making in asset $$1$$, relative to its cost. Imposing that $$\gamma <1$$ then ensures that the government does not implausibly trade unlimited amounts of asset $$1$$ in pursuit of $$p_{1,1}^{T}$$. This feature of Equation (4) is further discussed in Section 1.2.1. At date $$t=1$$, MMs in each asset $$i$$ clear their market after observing its aggregate order flow, $$\omega _{i}$$, like in Section 1.1. However, while $$\omega _{2}=\sum_{m=1}^{M}x_{2}\left( m\right) +z_{2}$$, $$\omega _{1}$$ now consists of the market orders of noise traders, speculators, and the government: $$\omega _{1}=x_{1}\left({\it gov}\right) +\sum_{m=1}^{M}x_{1}\left( m\right) +z_{1}$$. In this amended economy, MMs in each asset $$i$$ attempt to learn from $$\omega _{i}$$ about that asset’s terminal payoff $$v$$ when setting its equilibrium price $$p_{1,i}$$, like in Section 1.1. However, each speculator now uses her private signal, $$S_{v}\left( m\right) $$, to learn not only about $$v$$ and the other speculators’ private signals but also about the government’s intervention policy in asset $$1$$ before choosing her optimal trading strategy, $$x_{i}\left( m\right) $$, in both assets $$1$$ and $$2$$. In addition, the government uses its private information, $$S_{v}\left({\it gov}\right) $$, to learn about what speculators may know about $$v$$ and trade in asset 1 when choosing its optimal intervention strategy, $$x_{1}\left({\it gov}\right) $$. I solve for the ensuing unique linear Bayesian Nash equilibrium in Proposition 2. Proposition 2. There exists a unique linear equilibrium given by the price functions: \begin{align} p_{1,1}^{\ast } &=\left[ p_{0}+2d\lambda ^{\ast }\left( p_{0}-\overline{p}_{1,1}^{T}\right) \right] +\lambda ^{\ast }\omega _{1},\label{price1} \\ \end{align} (5) \begin{align} p_{1,2}^{\ast } &=p_{0}+\lambda \omega _{2}\text{,} \label{price2} \end{align} (6)where$$d=\frac{\gamma }{1-\gamma }$$, $$\lambda ^{\ast }$$ is the unique positive real root of the sextic polynomial of Equation (A33) in the Appendix, and $$\lambda =\frac{\sigma _{v}\sqrt{M\rho }}{\sigma _{z}\left[ 2+\left( M-1\right) \rho \right] }>0$$ (like in Proposition 1); by each speculator’s orders \begin{align} x_{1}^{\ast }\left( m\right) &=B_{1,1}^{\ast }\delta _{v}\left( m\right) \text{,} \label{trade1} \\ \end{align} (7) \begin{align} x_{2}^{\ast }\left( m\right) &=\frac{\sigma _{z}}{\sigma _{v}\sqrt{M\rho }}\delta _{v}\left( m\right), \label{trade2} \end{align} (8) where $$B_{1,1}^{\ast }=\frac{2-\psi }{\lambda ^{\ast }\left\{ 2\left[ 2+\left( M-1\right) \rho \right] \left( 1+d\lambda ^{\ast }\right) -M\rho \psi \left( 1+2d\lambda ^{\ast }\right) \right\} }>0$$; and by the government intervention: \begin{equation} x_{1}\left({\it gov}\right) =2d\left( \overline{p}_{1,1}^{T}-p_{0}\right) +C_{1,1}^{\ast }\delta _{v}\left({\it gov}\right) +C_{2,1}^{\ast }\delta _{T}\left({\it gov}\right) \text{,} \label{tradegov} \end{equation} (9) where $$C_{1,1}^{\ast }=\frac{\left[ 2+\left( M-1\right) \rho \right] \left(1+d\lambda ^{\ast }\right) -M\rho \left( 1+2d\lambda ^{\ast }\right) }{\lambda ^{\ast }\left( 1+d\lambda ^{\ast }\right) \left\{ 2\left[ 2+\left( M-1\right) \rho \right] \left( 1+d\lambda ^{\ast }\right) -M\rho \psi \left(1+2d\lambda ^{\ast }\right) \right\} }$$ and $$C_{2,1}^{\ast }=\frac{d}{1+d\lambda ^{\ast }}>0$$. In Corollary 3, I examine the effect of government intervention in asset $$1$$, $$x_{1}\left({\it gov}\right)$$ of Equation (9), on the extent of LOP violations in the economy (i.e., on the unconditional comovement of equilibrium asset prices $$p_{1,1}^{\ast }$$ and $$p_{1,2}^{\ast}$$ of Equations (5) and (6)), like in Corollary 1. Corollary 3. In the presence of government intervention, the unconditional correlation of the equilibrium prices of assets $$1$$ and $$2$$ is given by: \begin{equation} {\it corr}( p_{1,1}^{\ast },p_{1,2}^{\ast }) =\frac{\sigma _{zz}+\sigma _{z}\sigma _{v}\sqrt{M\rho }\{ B_{1,1}^{\ast }[ 1+( M-1) \rho ] +\psi C_{1,1}^{\ast }+C_{2,1}^{\ast }\} }{\sigma _{z}\sqrt{[ 2+( M-1) \rho ] \{ \sigma _{z}^{2}+\sigma _{v}^{2}\{ M\rho B_{1,1}^{\ast 2}[ 1+(M-1) \rho ] +D_{1}^{\ast }+E_{1}^{\ast }\} \} }}, \label{correlationgov} \end{equation} (10) where $$D_{1}^{\ast }=2M\rho \left[ B_{1,1}^{\ast }\left( \psi C_{1,1}^{\ast }+C_{2,1}^{\ast }\right) \right] $$ and $$E_{1}^{\ast }=\psi C_{1,1}^{\ast 2}+\frac{1}{\mu \psi }C_{2,1}^{\ast 2}+2C_{1,1}^{\ast }C_{2,1}^{\ast }$$. There are no LOP violations under perfectly integrated market making. In the above economy, the equilibrium price impact of order flow in asset $$1$$ ($$\lambda ^{\ast }$$ of Proposition 2) cannot be solved in closed form (see the Appendix). Thus, I characterize the equilibrium properties of $${\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right)$$ of Equation (10) via numerical analysis. To that purpose, I introduce the stylized government, with starting parameters $$\gamma =0.5$$, $$\psi =0.5$$, and $$\mu =0.5$$, in the baseline economy of Section 1.1.2, where $$\sigma _{v}^{2}=1$$, $$\sigma _{z}^{2}=1$$, $$\sigma _{zz}=0.5$$, $$\rho =0.5$$, and $$M=10$$. Most parameter selection only affects the relative magnitude of the effects described below. I examine limiting cases and nonrobust exceptions of interest in Section 1.2.1; see also the discussion in the proof of Proposition 2. I then plot the ensuing equilibrium price correlation $${\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) $$ in Figure 1 (dashed lines), alongside its corresponding level in the absence of government intervention ($${\it corr}\left( p_{1,1},p_{1,2}\right) $$ of Equation (3), solid lines), as a function of $$\sigma _{zz}$$, $$\rho $$, $$M$$, or $$\sigma _{z}^{2}$$ (Figures 1A to 1D, like in Section 1.1.2), and $$\gamma $$, $$\mu $$, $$\psi $$, or $$\sigma _{v}^{2}$$ (Figure 1E to 1H). Figures 2B to 2C display their difference, $$\Delta {\it corr}\left( p_{1,1},p_{1,2}\right) \equiv {\it corr}\left( p_{1,1},p_{1,2}\right) -{\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) $$, as a function of the corresponding average $$\lambda $$ (i.e., $$\overline{\lambda }=\frac{1}{2}\left( \lambda +\lambda ^{\ast }\right) $$) and $${\it corr}\left( p_{1,1},p_{1,2}\right) $$ (i.e., $$\overline{{\it corr}\left( p_{1,1},p_{1,2}\right) }=\frac{1}{2}\left[ {\it corr}\left( p_{1,1},p_{1,2}\right) +{\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) \right] $$), respectively, for both their relation with $$\rho $$ of Figure 1B (solid line, right axis, for $$\sigma _{z}^{2}=1$$ and $$\rho \approx 0.5$$) and their relation with $$\sigma _{z}^{2}$$ of Figure 1D (dashed line, left axis, for $$\rho =0.5$$ and $$\sigma _{z}^{2}\approx 1$$). Like in Corollary 1, if MMs observe order flow in both assets $$1$$ and $$2$$, once again no LOP violation can arise in equilibrium under semi-strong market efficiency, regardless of government intervention: $${\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) ={\it corr}\left( p_{1,1},p_{1,2}\right) =1$$. However, insofar as the dealership sector is segmented and multiasset speculators submit market orders (i.e., ceteris paribus for existing limits to arbitrage), government intervention makes LOP violations more likely in equilibrium, even in the absence of liquidity demand differentials. According to Figure 1, official trading activity in asset $$1$$ lowers the unconditional correlation of the equilibrium prices of the otherwise identical assets $$1$$ and $$2$$ (i.e., $${\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) <{\it corr}\left( p_{1,1},p_{1,2}\right) $$) even when noise trading in those assets is perfectly correlated (i.e., $$\sigma _{zz}=\sigma _{z}^{2}=1$$ such that $${\it corr}\left( p_{1,1},p_{1,2}\right) =1$$ in Figure 1A). Intuitively, the camouflage provided by the aggregate order flow allows the stylized government of Equation (4) to trade in asset $$1$$ to push its equilibrium price $$p_{1,1}^{\ast }$$ toward a target $$p_{1,T}$$ that is at most only partially informative about fundamentals, that is, only partially correlated with both assets’ identical terminal payoff $$v$$: $${\it corr}\left( v,p_{1,1}^{T}\right) =\sqrt{\mu \psi }<1$$ (see also Vitale 1999; Naranjo and Nimalendran 2000). To that end, the government optimally chooses to bear some costs, that is, to tolerate some trading losses or forego some trading profits in asset $$1$$, given its private information of precision $$\psi $$. For instance, at the economy’s baseline parametrization, not only is $$C_{2,1}^{\ast }>0$$ but also $$0<C_{1,1}^{\ast }<B_{1,1}^{\ast }$$ in $$x_{1}\left({\it gov}\right) $$ of Equation (9): $$C_{2,1}^{\ast }=0.85 $$ and $$C_{1,1}^{\ast }=0.34$$ versus $$B_{1,1}^{\ast }=0.69$$ in $$x_{1}^{\ast }\left( m\right) $$ of Equation (7). Since $$p_{1,1}^{T}$$ is also nonpublic (i.e., policy uncertainty $$\sigma _{T}^{2}=\frac{\sigma _{v}^{2}}{\mu \psi }>0$$), the uninformed MMs in asset $$1$$ cannot fully account for the government’s trading activity when setting $$p_{1,1}^{\ast }$$ from the observed aggregate order flow in that asset, $$\omega _{1}$$ (i.e., $$E\left( v|\omega _{1}\right) $$). As such, camouflaged government intervention in asset $$1$$ is at least partly effective at pushing that asset’s equilibrium price $$p_{1,1}^{\ast }$$ toward its partly uninformative policy target $$p_{1,1}^{T}$$—ceteris paribus, $$\frac{\partial p_{1,1}^{\ast }}{\partial p_{1,1}^{T}}=\frac{d\lambda ^{\ast }}{1+d\lambda ^{\ast }}>0$$ in Proposition 2—hence away from the equilibrium price of asset $$2$$, $$p_{1,2}^{\ast }$$, despite occurring in a deeper market. For instance, in the baseline economy, $$\lambda ^{\ast }=0.18$$ versus $$\lambda =0.34$$. Intuitively, $$\lambda ^{\ast }<\lambda $$ because at least partly uninformative official trading activity in asset $$1$$ both alleviates dealers’ adverse selection risk and induces more aggressive informed (i.e., perfectly correlated) speculation in that asset (Subrahmanyam 1991b; Pasquariello, Roush, and Vega 2014): $$B_{1,1}^{\ast }>\frac{\sigma _{z}}{\sigma _{v}\sqrt{M\rho }}$$ in Equations (7) and (8), respectively; for example, $$B_{1,1}^{\ast }=0.69$$ versus $$\frac{\sigma _{z}}{\sigma _{v}\sqrt{M\rho }}=0.45$$. This liquidity differential mitigates the differential impact of less-than-perfectly correlated noise trading shocks on $$p_{1,1}^{\ast }$$ and $$p_{1,2}^{\ast }$$.14 However, ceteris paribus for $$p_{1,2}^{\ast }$$, the former effect of government intervention on $$p_{1,1}^{\ast }$$ prevails on its latter effect on asset $$1$$’s liquidity, leading to greater LOP violations in equilibrium (i.e., allowing for further $$E\left( v|\omega _{1}\right) \neq E\left( v|\omega _{2}\right) $$). For instance, in the baseline economy, $${\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) =0.89$$ versus $${\it corr}\left(p_{1,1},p_{1,2}\right) =0.92$$, which amounts to a $$19\%$$ increase in the expected absolute difference between $$p_{1,1}$$ and $$p_{1,2}$$, $$E\left( \left\vert p_{1,1}-p_{1,2}\right\vert \right) $$.15 Consistently, so-induced LOP violations increase (lower $${\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) $$) not only when the government is more committed to achieve its policy target $$p_{1,1}^{T}$$ for asset $$1$$ (higher $$\gamma $$, Figure 1E), but also when the target is less correlated to its private signal of $$v$$, $$S_{v}\left({\it gov}\right) $$ (lower $$\mu $$, Figure 1F), or that signal is less precise (lower $$\psi $$, Figure 1G) such that its official trading activity in that asset is more costly yet less predictable. I further investigate this trade-off in Section 1.2.1. The implications of government intervention for LOP violations also depend on extant market conditions. Figures 1 and 2 suggest that official trading activity leads to larger LOP violations when the affected markets are less liquid and LOP violations are more severe in the government’s absence. In particular, equilibrium $${\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) $$ is lower (and lower than $${\it corr}\left( p_{1,1},p_{1,2}\right) $$) in the presence of fewer speculators (lower $$M$$, Figure 1C) or when their private information is more dispersed (lower $$\rho $$, Figure 1B and Figures 2B to 2C [solid lines]). Ceteris paribus, as discussed in Section 1.1.1, fewer or more heterogeneous speculators trade (as a group or individually) more cautiously with their private signals, making MMs’ adverse selection problem more severe and the equilibrium price impact of order flow (Kyle (1985) lambda) higher in both assets $$1$$ ($$\lambda $$) and $$2$$ ($$\lambda ^{\ast }$$), thereby lowering liquidity in both markets and amplifying the impact of liquidity demand differentials on their price correlation. In those circumstances, government intervention in asset $$1$$ is more effective at driving its equilibrium price $$p_{1,1}^{\ast }$$ of Equation (5) toward the partially uninformative policy target, $$p_{1,1}^{T}$$—ceteris paribus, $$\frac{\partial ^{2}p_{1,1}^{\ast }}{\partial p_{1,1}^{T}\partial \lambda ^{\ast }}=\frac{d}{\left( 1+d\lambda ^{\ast }\right) ^{2}}>0$$—hence farther away from the equilibrium price of asset $$2$$ ($$p_{1,2}^{\ast }$$ of Equation (6)). This effect, however, is less pronounced in correspondence with greater fundamental uncertainty (higher $$\sigma _{v}^{2}$$, Figure 1H). When private fundamental information is more valuable, both market liquidity deteriorates (see Section 1.1.1) and the pursuit of policy motives becomes more costly for the government in the loss function of Equation (4). The latter partly offsets the former, leading to a nearly unchanged $${\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) $$. Similarly, Figures 1 and 2 also suggest that government intervention may amplify LOP violations more conspicuously (greater $$\Delta {\it corr}\left( p_{1,1},p_{1,2}\right) >0$$) even when those violations are not as severe in its absence (high $${\it corr}\left( p_{1,1},p_{1,2}\right) $$). This may occur when noise trading in assets $$1$$ and $$2$$ ($$z_{1}$$ and $$z_{2}$$) is less intense, lowering liquidity in both markets (lower $$\sigma _{z}^{2}$$, Figure 1B and Figures 2B to 2C [dashed lines]), or when $$z_{1}$$ and $$z_{2}$$ are more positively correlated (higher $$ \sigma _{zz}$$, Figure 1A). For instance, in the baseline economy with perfectly correlated noise trading shocks ($$\sigma _{zz}=\sigma _{z}^{2}=1$$ ), $${\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) =0.93$$ (and $$E\left( \left\vert p_{1,1}^{\ast }-p_{1,2}^{\ast }\right\vert \right) =0.27$$) versus $${\it corr}\left( p_{1,1},p_{1,2}\right) =1$$ (and $$E\left( \left\vert p_{1,1}-p_{1,2}\right\vert \right) =0$$). Hence, the observed relation between the impact of government intervention on LOP violations and their extant severity may be positive, negative, or possibly nonmonotonic. I summarize these novel, robust observations about the impact of government intervention on the LOP in Conclusions 1 and 2.16 Conclusion 1. Under less-than-perfectly integrated market making, government intervention results in greater LOP violations in equilibrium, even in the absence of liquidity demand differentials. Conclusion 2. Government-induced LOP violations increase in the government’s policy commitment, speculators’ information heterogeneity, policy (but not fundamental) uncertainty, and the covariance of noise trading, as well as decrease in the quality of the government’s private fundamental information, the covariance of its policy target with fundamentals, the number of speculators, and the intensity of noise trading. 1.2.1 Limiting cases and exceptions In this section, I examine the implications of notable limiting cases of the model of Section 1.2 for the positive relation between government intervention and LOP violations postulated in Conclusion 1. All of these circumstances are arguably less plausible relative to the aforementioned literature on official trading activity, and some of them may yield nonrobust exceptions to Conclusion 1. Yet, their examination allows me to further illustrate the intuition behind the model’s main predictions. To begin with, if $$\gamma =0$$ in the loss function of Equation (4), the government in the model would act exclusively as an additional, privately informed trader in asset $$1$$. The equilibrium of the resulting economy can be shown to closely mimic the one in Proposition 1 except in that such intervention would make only asset $$1$$ both more liquid ($$\lambda ^{\ast }<\lambda $$) and more informationally efficient ($$var\left( p_{1,1}^{\ast }\right) >var\left( p_{1,1}\right) $$), like by increasing the total number of speculators $$M$$ by one unit only in asset $$1$$ (see Section 1.1.1), and especially when $$M$$ is small; thus, it would lower asset $$1$$’s equilibrium price correlation with asset $$2$$ relative to Corollary 1 ($${\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) <{\it corr}\left( p_{1,1},p_{1,2}\right) $$), even in the presence of perfectly correlated noise trading shocks ($$\sigma _{zz}=\sigma _{z}^{2}$$). See, for example, Figure IA-1A in the Internet Appendix. The equilibrium $${\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) $$ of Corollary 3 and Figure 1E converges to this limiting case for $$\gamma \rightarrow 0$$. Relatedly, there are also circumstances when the dispersion of the information endowments of a sufficiently small number of speculators is so high (i.e., when the precision and correlation of their private signals of $$v$$ are so low, $$\rho \approx 0$$) that the government is practically the only informed trader in the targeted asset, thus worsening its dealers’ adverse selection risk such that $$\lambda ^{\ast }>\lambda $$ (e.g., like in Vitale 1999; Naranjo and Nimalendran 2000) and $${\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) \ll {\it corr}\left(p_{1,1},p_{1,2}\right) $$, like in Conclusion 1. Conclusion 1 is also robust to imposing that the government’s policy target $$p_{1,T}$$ is independent of asset $$1$$’s terminal payoff $$v$$ (i.e., $$cov\left( v,p_{1,1}^{T}\right) =0$$, like in Pasquariello, Roush, and Vega 2014), or when $$\mu \rightarrow 0$$ such that $${\it corr}\left( v,p_{1,1}^{T}\right) =\sqrt{\mu \psi }\rightarrow 0$$. See, for example, Figure IA-1B in the Internet Appendix. This is true even if the government is uninformed about asset fundamentals, that is, even in the absence of $$S_{v}\left({\it gov}\right) $$, or when $$\psi \rightarrow 0$$ such that $${\it corr}\left[ v,S_{v}\left({\it gov}\right) \right] =\sqrt{\psi }\rightarrow 0$$. Intuitively, in either case the pursuit of policy may be not only more costly for the government in terms of expected trading losses in asset $$1$$, but also more effective as less predictable to other market participants. It can be shown that the equilibrium $${\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) $$ of Corollary 3 and Figures 1F to 1G converges to either of these limiting cases for either $$\mu \rightarrow 0$$ (but $$\psi >0$$) or $$\mu =\psi \rightarrow 0$$, respectively. Relatedly, there are also circumstances when an informed government may optimally trade in asset $$1$$ against its private information (“leaning against the wind”) to achieve its at least partly informative policy objectives. For instance, consider parametrizations of the baseline economy for which the equilibrium price impact of order flow in either asset $$1$$ or $$2$$ is relatively low (e.g., $$\rho =0.9$$ such that $$\lambda =0.29$$) and the government’s price target is both relatively important in its loss function ($$\gamma =0.5$$ in $$ L\left({\it gov}\right) $$ of Equation (4)) and only partially correlated to its fundamental information ($$\mu =0.5$$ such that $${\it corr}\left[ p_{1,1}^{T},S_{v}\left({\it gov}\right) \right] =\sqrt{\mu }=0.71$$). In such economies, the resulting $$C_{1,1}^{\ast }<0$$ in $$x_{1}\left({\it gov}\right) $$ of Equation (9), while $$B_{1,1}^{\ast }>0$$ in $$x_{1}^{\ast }\left( m\right) $$ of Equation (7): $$C_{1,1}^{\ast }=-0.04$$ versus $$B_{1,1}^{\ast }=0.55$$. Lastly, government intervention in asset $$1$$ may reduce LOP violations in equilibrium when $$\sigma _{zz}$$ is close to zero or negative (such that liquidity trading in the fundamentally identical assets $$1$$ and $$2$$ is weakly or negatively correlated), or when both $$\psi $$ and $$\mu $$ are close to one (such that a nearly fully informed government is in pursuit of a nearly fully informative policy target). In those more extreme circumstances—but only under some market conditions, like a relatively large number of speculators, and even if the government is uninformed and/or in pursuit of an uninformative target—such intervention may increase equilibrium price correlation ($${\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) >{\it corr}\left( p_{1,1},p_{1,2}\right) $$), in exception to Corollary 1, by at least partly offsetting the impact of highly divergent noise trading shocks on $$p_{1,1}^{\ast }$$. See, for example, Figure IA-1C in the Internet Appendix. 1.3 Empirical implications The stylized model of Sections 1.1 and 1.2 represents a plausible channel through which direct government intervention may affect the relative prices of fundamentally linked securities in markets with less-than-perfectly integrated dealership. This channel depends crucially on various facets of government policy and the information environment of those markets. Yet, measuring such intervention characteristics and market conditions is challenging, and often unfeasible. Under these premises, I identify from Corollary 1, Proposition 2, Figures 1 and 2, and Conclusions 1 and 2 the following subset of plausibly testable implications of official trading activity for relative mispricings: H1 government intervention does not affect extant LOP violations, if any, in markets with perfectly integrated dealership; H2 government intervention induces, or increases extant LOP violations in markets with less-than-perfectly integrated dealership; H3 this effect is more pronounced when market liquidity is low; H4 this effect is more pronounced when information heterogeneity is high; and H5 this effect is more pronounced when government policy uncertainty is high. 2. Empirical Analysis I test the implications of my model by analyzing the impact of government intervention in currency markets on the relative pricing of American Depositary Receipts and other U.S. cross-listings (“ADRs” for brevity). An ADR is a dollar-denominated security, traded in the United States, representing ownership of a pre-specified amount (“bundling ratio”) of stocks of a foreign company, denominated in a foreign currency, held on deposit at a U.S. depositary banks (e.g., Karolyi 1998, 2006). In Section 2.1, I motivate the use of this setting to that purpose. I describe the data in Section 2.2.Sections 2.3 to 2.5 contain the econometric analysis. 2.1 ADRs and forex intervention in the model The market for U.S. cross-listings (the “ADR market”) represents an ideal setting to test my model, since its interaction with the foreign exchange (“forex”) market is consistent in spirit with the model’s basic premises. First, exchange rates and ADRs are fundamentally linked by an arbitrage parity. Depositary banks facilitate the convertibility between ADRs and their underlying foreign shares (Gagnon and Karolyi 2010) such that the unit price of an ADR $$i$$, $$P_{i,t}$$, should at any time $$t$$ be equal to the dollar (USD) price of the corresponding amount (bundling ratio) $$q_{i}$$ of foreign shares, $$P_{i,t}^{\it LOP}$$: \begin{equation} P_{i,t}^{\it LOP}=S_{t,USD/FOR}\times q_{i}\times P_{i,t}^{\it FOR}\text{,} \label{ADR} \end{equation} (11) where $$P_{i,t}^{\it FOR}$$ is the unit foreign stock price denominated in a foreign currency FOR, and $$S_{t,USD/FOR}$$ is the exchange rate between USD and FOR. I interpret the fundamental commonality in the terminal payoffs of assets $$1$$ and $$2$$ in the model ($$v_{1}$$ and $$v_{2}$$) as a stylized representation of the LOP relation between currency and ADR markets in Equation (11). In particular, Equation (11) suggests that one can think of asset $$1$$ as the exchange rate—with payoff $$v_{1}=v$$—traded in the forex market at a price $$p_{1,1}$$ (i.e., $$S_{t,USD/FOR}$$); and of asset $$2$$ as an ADR—whose payoff $$v_{2}$$ is a linear function of the exchange rate: $$v_{2}=a_{2}+b_{2}v$$, where $$a_{2}=0$$ and $$b_{2}=q_{i}\times P_{i,t}^{\it FOR}>0$$, that is, ceteris paribus for the corresponding foreign stock price—traded in the U.S. stock market at a tilded price $$\widetilde{p}_{1,2}=b_{2}p_{1,2}$$ (i.e., $$P_{i,t}$$). Ignoring the market for an ADR’s underlying foreign shares is for simplicity only and without loss of generality. In Section 1 and Figure IA-2 of the Internet Appendix, I show that extending the model to a third such asset—with payoff $$v_{3}$$ such that the ADR’s log-linearized payoff $$v_{2}=a_{2}+v_{1}+v_{3}$$, where $$a_{2}=\ln \left( q_{i}\right) $$—requires more involved analysis but yields similar implications. In the above setting, the LOP relation between actual ($$P_{i,t}$$) and synthetic ($$P_{i,t}^{\it LOP}$$) ADR prices in Equation (11) can then be represented by the unconditional correlation between $$\widetilde{p}_{1,2}$$ and $$p_{1,2}^{\it LOP}=b_{2}p_{1,1}$$, respectively (e.g., Gromb and Vayanos 2010), such that in equilibrium: $${\it corr}\left( \widetilde{p} _{1,2},p_{1,2}^{\it LOP}\right) ={\it corr}\left( p_{1,1},p_{1,2}\right) $$ of Equation (3). Accordingly, I postulate in Conclusion 1 that, ceteris paribus, government intervention in the forex market—that is, targeting the exchange rate $$p_{1,1}$$—lowers the unconditional correlation between exchange rates and actual ADR prices—that is, between $$p_{1,1}$$ and $$\widetilde{p}_{1,2}$$: $${\it corr}\left( p_{1,1}^{\ast },\widetilde{p} _{1,2}^{\ast }\right) ={\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) $$ of Equation (10), such that $${\it corr}\left( p_{1,1}^{\ast }, \widetilde{p}_{1,2}^{\ast }\right) <{\it corr}\left( \widetilde{p}_{1,2},p_{1,2}^{\it LOP}\right) $$. Hence, forex intervention may yield larger price differentials between actual and synthetic ADRs—that is, it lowers the unconditional correlation between $$\widetilde{p}_{1,2}$$ and $$ p_{1,2}^{\it LOP}$$: $${\it corr}\left( \widetilde{p}_{1,2}^{\ast },p_{1,2}^{LOP\ast }\right) ={\it corr}\left( p_{1,1}^{\ast },p_{1,2}^{\ast }\right) $$, such that $$ {\it corr}\left( \widetilde{p}_{1,2}^{\ast },p_{1,2}^{LOP\ast }\right) <{\it corr}\left( \widetilde{p}_{1,2},p_{1,2}^{\it LOP}\right) $$. Second, market making in currency and ADR markets is arguably less-than-perfectly integrated, in that market makers in one market are less likely to directly observe, and set prices based on, trading activity in the other market than within their own.17 I interpret segmented market making in assets $$1$$ and $$2$$ in the model as a stylized representation of this observation. Third, as mentioned in Section 1.2, the stylized representation of the government in the model is consistent with the consensus in the literature that government intervention in currency markets, although typically secret and in pursuit of nonpublic policy, is often effective at moving exchange rates because it is deemed at least partly informative about fundamentals.18 Fourth, the same literature suggests that forex intervention is unlikely to be motivated by relative mispricings in the ADR market. This observation alleviates reverse causality concerns when estimating and interpreting any empirical relation between government intervention and the arbitrage parity of Equation (11). I further assess this and other potential sources of endogeneity in Section 2.3.1. Overall, according to the model, these features of currency and ADR markets raise the possibility that government intervention in the former may lead to LOP violations in the latter, that is, to “ADR parity” (ADRP) violations. I measure these violations as nonzero absolute log percentage differences, in basis points (bps), between actual ($$P_{i,t}$$) and theoretical ADR prices ($$P_{i,t}^{\it LOP}$$ of Equation (11)): \begin{equation} ADRP_{i,t}=\left\vert \ln \left( P_{i,t}\right) -\ln \left( P_{i,t}^{\it LOP}\right) \right\vert \times 10,000 \label{parity} \end{equation} (12) (e.g., Gagnon and Karolyi 2010; Pasquariello, Roush, and Vega 2014), and assess their empirical relation with forex intervention in the reminder of the paper. 2.1.1 Alternative model interpretations and measures of ADRP violations My investigation of the effects of forex interventions on ADRP violations is qualitatively unaffected when considering alternative interpretations of the traded assets in the model, relative to actual and synthetic ADRs in Equation (11), or alternative measures of LOP violations both in the model and in the ADR market, relative to their absolute price differentials in Equation (12). To begin with, I show in Section 2 of the Internet Appendix that the linearity of asset payoffs and equilibrium prices in the model implies that one can also think of asset $$1$$ as the actual exchange rate traded in the forex markets and of asset $$2$$ as either: (1) an ADR-specific synthetic, or shadow exchange rate implied by Equation (11) implicitly traded in the ADR market at $$S_{\it t,USD/FOR}^{\it i,LOP}=P_{i,t}\times \left( q_{i}\times P_{i,t}^{\it FOR}\right) ^{-1}$$ (e.g., Auguste et al. 2006; Eichler, Karmann, and Maltritz 2009); or (2) an actual ADR traded in the U.S. stock market at $$ P_{i,t}$$ implying a synthetic exchange rate $$S_{\it t,USD/FOR}^{i,LOP}$$. Although less common and intuitive, these representations of the LOP relation between currency and ADR markets are conceptually and empirically equivalent to the one discussed in Section 2.1 since any violation of the ADR parity of Equation (11) yields both $$P_{i,t}\neq P_{i,t}^{\it LOP}$$ and $$S_{\it t,USD/FOR}\neq S_{\it t,USD/FOR}^{i,LOP}$$—that is, not only the same equilibrium price correlation in the model but also the same absolute percentage LOP violation in Equation (12). In addition, as noted in Section 1.1.2, the notion of LOP violations in the ADR market as nonzero unsigned relative, that is, log percentage, price differentials $${\it ADRP}_{i,t}$$ of Equation (12) is both common in the literature and conceptually equivalent to the notion of LOP violations as less-than-one equilibrium unconditional price correlation $${\it corr}\left(p_{1,1},p_{1,2}\right) $$ in the model. For instance, I show in Section 3 of the Internet Appendix that the expected absolute differential between equilibrium actual and synthetic ADR prices described in Section 2.1 (i.e., $$E\left( \left\vert \widetilde{p}_{1,2}-p_{1,2}^{\it LOP}\right\vert \right) $$) is a, ceteris paribus decreasing, function of their unconditional correlation whose scale depends on the magnitude of the ADR’s fundamental payoff. Both $${\it corr}\left( p_{1,1},p_{1,2}\right) $$ and $${\it ADRP}_{i,t}$$ are instead price-scale invariant and display similar comparative statics (see also Auguste et al. 2006; Pasquariello 2008; Gagnon and Karolyi 2010). Accordingly, the empirical analysis of several measures of the correlation between actual and synthetic ADR prices, although computationally less convenient than for $${\it ADRP}_{i,t}$$ in my setting, yields qualitatively similar inference. See, for example, Figure IA-3 and Tables IA-1 and IA-2 in the Internet Appendix. 2.2 Data For the empirical investigation of my model, I construct a sample of ADRs traded in U.S. stock exchanges and official intervention activity in currency markets over the past three decades. 2.2.1 ADRs I begin by obtaining from Thomson Reuters Datastream (Datastream) its entire sample of foreign stocks cross-listed in the United States between January 1, 1973 and December 31, 2009.19 Following standard practice in the literature, I then remove ADRs trading over-the-counter (Level I), Securities and Exchange Commission (SEC) Regulation S shares, private placement ADRs (Rule 144A), and preferred shares. In addition, I also conservatively exclude any identifiable cross-listing with ambiguous, incomplete, or missing descriptive, listing, or pairing information in the Datastream sample. This leaves a subset of $$410$$ viable Level II and Level III ADRs from developed and emerging countries (with bundling ratios $$q_{i}$$) and mostly Canadian ordinary shares (ordinaries, with $$q_{i}=1$$) listed on the three major U.S. stock exchanges (NYSE, AMEX, or NASDAQ).20 Because of my focus on forex interventions, Table 1 reports the composition of this sample by the country or most recent currency area of listing (i.e., most recent currency of denomination) of the underlying foreign stocks. Most viable cross-listed stocks are traded in developed, highly liquid and higher-quality equity markets, and denominated in highly liquid currencies: Canada (CAD, $$N_{v}=67$$), Euro area (EUR, $$58$$), United Kingdom (GBP, $$43$$), Australia (AUD, $$30$$), and Japan (JPY, $$24$$); emerging, often less liquid and lower-quality equity markets and currencies of local listing comprise Hong Kong (HKD, $$54$$ including H-shares of firms incorporated in mainland China), Brazil (BRL, $$23$$), and South Africa (ZAR, $$14$$), among others. Table 1 ADRP violations: Summary statistics ADRP violations ADRP illiquidity $${\it ADRP}_{m}$$ $${\it ADRP}_{m}^{z}$$ $$\Delta {\it ADRP}_{m}$$ $$\Delta {\it ADRP}_{m}^{z}$$ $${\it ILLIQ}_{m}$$ Country (Currency) $$N_{v}$$ $$N_{u}$$ $$N$$ Mean SD Mean SD Mean SD Mean SD Mean SD Australia (AUD) 30 23 269 217.26 112.57 $$-$$0.23 0.32 $$-$$3.39 53.91 $$-$$0.002 0.280 11.0% 9.0% Argentina (ARS) 9 6 198 187.70 114.41 0.40 1.15 0.44 81.45 $$-$$0.001 0.840 19.2% 12.5% Brazil (BRL) 23 18 178 165.06 104.49 $$-$$0.08 0.36 $$-$$1.60 40.57 0.003 0.307 9.1% 7.2% Canada (CAD) 67 46 360 103.13 46.08 $$-$$0.18 0.27 $$-$$0.08 26.14 $$-$$0.000 0.189 13.4% 7.7% Chile (CLP) 5 5 168 185.16 67.26 $$-$$0.33 0.38 0.41 50.94 $$-$$0.006 0.338 14.9% 7.1% Euro area (EUR) 58 51 287 352.29 287.90 $$-$$0.41 0.64 $$-$$0.87 60.98 $$-$$0.001 0.421 6.5% 3.8% Hong Kong (HKD) 54 39 198 166.78 65.22 0.07 0.33 $$-$$0.69 40.03 $$-$$0.002 0.282 15.0% 5.9% India (INR) 10 9 143 248.34 141.33 $$-$$0.07 0.38 $$-$$0.42 64.99 $$-$$0.004 0.359 4.7% 3.1% Indonesia (IDR) 5 2 168 181.19 79.67 $$-$$0.04 0.48 $$-$$0.93 67.90 $$-$$0.010 0.408 9.7% 4.9% Japan (JPY) 24 19 360 149.34 75.33 $$-$$0.04 0.42 $$-$$0.31 29.46 $$-$$0.001 0.289 9.3% 3.8% S. Korea (KRW) 8 7 141 328.73 187.75 $$-$$0.10 0.44 $$-$$0.86 74.53 $$-$$0.004 0.285 6.9% 4.4% Switzerland (CHF) 4 2 116 253.67 141.83 $$-$$0.55 0.39 $$-$$1.98 37.30 $$-$$0.015 0.261 4.1% 2.6% Turkey (TRY) 7 3 74 227.50 174.99 0.21 0.99 $$-$$1.30 104.33 $$-$$0.005 0.619 7.8% 4.0% United Kingdom (GBP) 43 34 360 200.59 73.82 $$-$$0.21 0.31 $$-$$0.48 34.07 0.000 0.211 4.7% 2.6% Other (Other) 33 26 250 261.15 112.06 $$-$$0.18 0.40 $$-$$0.26 85.10 0.005 0.371 17.3% 14.1% Total 410 319 360 194.33 41.34 $$-$$0.17 0.19 $$-$$0.28 21.47 $$-$$0.001 0.153 10.6% 3.3% ADRP violations ADRP illiquidity $${\it ADRP}_{m}$$ $${\it ADRP}_{m}^{z}$$ $$\Delta {\it ADRP}_{m}$$ $$\Delta {\it ADRP}_{m}^{z}$$ $${\it ILLIQ}_{m}$$ Country (Currency) $$N_{v}$$ $$N_{u}$$ $$N$$ Mean SD Mean SD Mean SD Mean SD Mean SD Australia (AUD) 30 23 269 217.26 112.57 $$-$$0.23 0.32 $$-$$3.39 53.91 $$-$$0.002 0.280 11.0% 9.0% Argentina (ARS) 9 6 198 187.70 114.41 0.40 1.15 0.44 81.45 $$-$$0.001 0.840 19.2% 12.5% Brazil (BRL) 23 18 178 165.06 104.49 $$-$$0.08 0.36 $$-$$1.60 40.57 0.003 0.307 9.1% 7.2% Canada (CAD) 67 46 360 103.13 46.08 $$-$$0.18 0.27 $$-$$0.08 26.14 $$-$$0.000 0.189 13.4% 7.7% Chile (CLP) 5 5 168 185.16 67.26 $$-$$0.33 0.38 0.41 50.94 $$-$$0.006 0.338 14.9% 7.1% Euro area (EUR) 58 51 287 352.29 287.90 $$-$$0.41 0.64 $$-$$0.87 60.98 $$-$$0.001 0.421 6.5% 3.8% Hong Kong (HKD) 54 39 198 166.78 65.22 0.07 0.33 $$-$$0.69 40.03 $$-$$0.002 0.282 15.0% 5.9% India (INR) 10 9 143 248.34 141.33 $$-$$0.07 0.38 $$-$$0.42 64.99 $$-$$0.004 0.359 4.7% 3.1% Indonesia (IDR) 5 2 168 181.19 79.67 $$-$$0.04 0.48 $$-$$0.93 67.90 $$-$$0.010 0.408 9.7% 4.9% Japan (JPY) 24 19 360 149.34 75.33 $$-$$0.04 0.42 $$-$$0.31 29.46 $$-$$0.001 0.289 9.3% 3.8% S. Korea (KRW) 8 7 141 328.73 187.75 $$-$$0.10 0.44 $$-$$0.86 74.53 $$-$$0.004 0.285 6.9% 4.4% Switzerland (CHF) 4 2 116 253.67 141.83 $$-$$0.55 0.39 $$-$$1.98 37.30 $$-$$0.015 0.261 4.1% 2.6% Turkey (TRY) 7 3 74 227.50 174.99 0.21 0.99 $$-$$1.30 104.33 $$-$$0.005 0.619 7.8% 4.0% United Kingdom (GBP) 43 34 360 200.59 73.82 $$-$$0.21 0.31 $$-$$0.48 34.07 0.000 0.211 4.7% 2.6% Other (Other) 33 26 250 261.15 112.06 $$-$$0.18 0.40 $$-$$0.26 85.10 0.005 0.371 17.3% 14.1% Total 410 319 360 194.33 41.34 $$-$$0.17 0.19 $$-$$0.28 21.47 $$-$$0.001 0.153 10.6% 3.3% This table reports the composition of the sample of U.S. cross-listings by the country or most recent currency area of listing (i.e., most recent currency of denomination) of the underlying foreign stocks, as well as summary statistics on the country-level and marketwide measures of their usable mispricings and illiquidity in the analysis. These measures are constructed by first obtaining all viable Level II and Level III ADRs and ordinaries (“ADRs”) listed on the NYSE, AMEX, or NASDAQ from the entire Datastream sample of U.S. cross-listings between January 1, 1973 and December 31, 2009; $$N_{v}$$ is their number in each grouping (where all those with available actual mispricings in the 1970s are usable afterward). $${\it ADRP}_{m}$$ and $${\it ADRP}_{m}^{z}$$ are then computed from Datastream and Pacific data as the monthly averages of daily equal-weighted means of available, filtered actual (in basis points [bps], i.e., multiplied by 10,000) and historically standardized absolute log violations of the ADR parity (ADRP; Equations (11) and (12)); $$\Delta {\it ADRP}_{m}={\it ADRP}_{m}-{\it ADRP}_{m-1}$$ and $$\Delta {\it ADRP}_{m}^{z}={\it ADRP}_{m}^{z}-{\it ADRP}_{m-1}^{z}$$. $${\it ILLIQ}_{m}$$ is a measure of ADRP illiquidity, defined in Section 2.2.1 as the equal-weighted mean (in percentage) of monthly averages of $$Z_{t}^{\it FOR}$$, $$Z_{t}$$, and $$Z_{t}^{FX}$$, the daily fractions of ADRs in $${\it ADRP}_{m}$$ whose underlying foreign stock, ADR, or exchange rate experiences a zero return on day $$t$$. For each grouping, I list the ensuing total number of usable ADRs ($$N_{u}$$), as well as available months ($$N$$), mean, and standard deviation for each measure over the sample period 1980–2009. The“Other” grouping includes Colombia (COP), Denmark (DKK), Egypt (EGP), Hungary (HUF), Israel (ILS), New Zealand (NZD), Norway (NOK), Philippines (PHP), Singapore (SGD), Sweden (SEK), Taiwan (TWD), Thailand (THB), and Venezuela (VEF). Table 1 ADRP violations: Summary statistics ADRP violations ADRP illiquidity $${\it ADRP}_{m}$$ $${\it ADRP}_{m}^{z}$$ $$\Delta {\it ADRP}_{m}$$ $$\Delta {\it ADRP}_{m}^{z}$$ $${\it ILLIQ}_{m}$$ Country (Currency) $$N_{v}$$ $$N_{u}$$ $$N$$ Mean SD Mean SD Mean SD Mean SD Mean SD Australia (AUD) 30 23 269 217.26 112.57 $$-$$0.23 0.32 $$-$$3.39 53.91 $$-$$0.002 0.280 11.0% 9.0% Argentina (ARS) 9 6 198 187.70 114.41 0.40 1.15 0.44 81.45 $$-$$0.001 0.840 19.2% 12.5% Brazil (BRL) 23 18 178 165.06 104.49 $$-$$0.08 0.36 $$-$$1.60 40.57 0.003 0.307 9.1% 7.2% Canada (CAD) 67 46 360 103.13 46.08 $$-$$0.18 0.27 $$-$$0.08 26.14 $$-$$0.000 0.189 13.4% 7.7% Chile (CLP) 5 5 168 185.16 67.26 $$-$$0.33 0.38 0.41 50.94 $$-$$0.006 0.338 14.9% 7.1% Euro area (EUR) 58 51 287 352.29 287.90 $$-$$0.41 0.64 $$-$$0.87 60.98 $$-$$0.001 0.421 6.5% 3.8% Hong Kong (HKD) 54 39 198 166.78 65.22 0.07 0.33 $$-$$0.69 40.03 $$-$$0.002 0.282 15.0% 5.9% India (INR) 10 9 143 248.34 141.33 $$-$$0.07 0.38 $$-$$0.42 64.99 $$-$$0.004 0.359 4.7% 3.1% Indonesia (IDR) 5 2 168 181.19 79.67 $$-$$0.04 0.48 $$-$$0.93 67.90 $$-$$0.010 0.408 9.7% 4.9% Japan (JPY) 24 19 360 149.34 75.33 $$-$$0.04 0.42 $$-$$0.31 29.46 $$-$$0.001 0.289 9.3% 3.8% S. Korea (KRW) 8 7 141 328.73 187.75 $$-$$0.10 0.44 $$-$$0.86 74.53 $$-$$0.004 0.285 6.9% 4.4% Switzerland (CHF) 4 2 116 253.67 141.83 $$-$$0.55 0.39 $$-$$1.98 37.30 $$-$$0.015 0.261 4.1% 2.6% Turkey (TRY) 7 3 74 227.50 174.99 0.21 0.99 $$-$$1.30 104.33 $$-$$0.005 0.619 7.8% 4.0% United Kingdom (GBP) 43 34 360 200.59 73.82 $$-$$0.21 0.31 $$-$$0.48 34.07 0.000 0.211 4.7% 2.6% Other (Other) 33 26 250 261.15 112.06 $$-$$0.18 0.40 $$-$$0.26 85.10 0.005 0.371 17.3% 14.1% Total 410 319 360 194.33 41.34 $$-$$0.17 0.19 $$-$$0.28 21.47 $$-$$0.001 0.153 10.6% 3.3% ADRP violations ADRP illiquidity $${\it ADRP}_{m}$$ $${\it ADRP}_{m}^{z}$$ $$\Delta {\it ADRP}_{m}$$ $$\Delta {\it ADRP}_{m}^{z}$$ $${\it ILLIQ}_{m}$$ Country (Currency) $$N_{v}$$ $$N_{u}$$ $$N$$ Mean SD Mean SD Mean SD Mean SD Mean SD Australia (AUD) 30 23 269 217.26 112.57 $$-$$0.23 0.32 $$-$$3.39 53.91 $$-$$0.002 0.280 11.0% 9.0% Argentina (ARS) 9 6 198 187.70 114.41 0.40 1.15 0.44 81.45 $$-$$0.001 0.840 19.2% 12.5% Brazil (BRL) 23 18 178 165.06 104.49 $$-$$0.08 0.36 $$-$$1.60 40.57 0.003 0.307 9.1% 7.2% Canada (CAD) 67 46 360 103.13 46.08 $$-$$0.18 0.27 $$-$$0.08 26.14 $$-$$0.000 0.189 13.4% 7.7% Chile (CLP) 5 5 168 185.16 67.26 $$-$$0.33 0.38 0.41 50.94 $$-$$0.006 0.338 14.9% 7.1% Euro area (EUR) 58 51 287 352.29 287.90 $$-$$0.41 0.64 $$-$$0.87 60.98 $$-$$0.001 0.421 6.5% 3.8% Hong Kong (HKD) 54 39 198 166.78 65.22 0.07 0.33 $$-$$0.69 40.03 $$-$$0.002 0.282 15.0% 5.9% India (INR) 10 9 143 248.34 141.33 $$-$$0.07 0.38 $$-$$0.42 64.99 $$-$$0.004 0.359 4.7% 3.1% Indonesia (IDR) 5 2 168 181.19 79.67 $$-$$0.04 0.48 $$-$$0.93 67.90 $$-$$0.010 0.408 9.7% 4.9% Japan (JPY) 24 19 360 149.34 75.33 $$-$$0.04 0.42 $$-$$0.31 29.46 $$-$$0.001 0.289 9.3% 3.8% S. Korea (KRW) 8 7 141 328.73 187.75 $$-$$0.10 0.44 $$-$$0.86 74.53 $$-$$0.004 0.285 6.9% 4.4% Switzerland (CHF) 4 2 116 253.67 141.83 $$-$$0.55 0.39 $$-$$1.98 37.30 $$-$$0.015 0.261 4.1% 2.6% Turkey (TRY) 7 3 74 227.50 174.99 0.21 0.99 $$-$$1.30 104.33 $$-$$0.005 0.619 7.8% 4.0% United Kingdom (GBP) 43 34 360 200.59 73.82 $$-$$0.21 0.31 $$-$$0.48 34.07 0.000 0.211 4.7% 2.6% Other (Other) 33 26 250 261.15 112.06 $$-$$0.18 0.40 $$-$$0.26 85.10 0.005 0.371 17.3% 14.1% Total 410 319 360 194.33 41.34 $$-$$0.17 0.19 $$-$$0.28 21.47 $$-$$0.001 0.153 10.6% 3.3% This table reports the composition of the sample of U.S. cross-listings by the country or most recent currency area of listing (i.e., most recent currency of denomination) of the underlying foreign stocks, as well as summary statistics on the country-level and marketwide measures of their usable mispricings and illiquidity in the analysis. These measures are constructed by first obtaining all viable Level II and Level III ADRs and ordinaries (“ADRs”) listed on the NYSE, AMEX, or NASDAQ from the entire Datastream sample of U.S. cross-listings between January 1, 1973 and December 31, 2009; $$N_{v}$$ is their number in each grouping (where all those with available actual mispricings in the 1970s are usable afterward). $${\it ADRP}_{m}$$ and $${\it ADRP}_{m}^{z}$$ are then computed from Datastream and Pacific data as the monthly averages of daily equal-weighted means of available, filtered actual (in basis points [bps], i.e., multiplied by 10,000) and historically standardized absolute log violations of the ADR parity (ADRP; Equations (11) and (12)); $$\Delta {\it ADRP}_{m}={\it ADRP}_{m}-{\it ADRP}_{m-1}$$ and $$\Delta {\it ADRP}_{m}^{z}={\it ADRP}_{m}^{z}-{\it ADRP}_{m-1}^{z}$$. $${\it ILLIQ}_{m}$$ is a measure of ADRP illiquidity, defined in Section 2.2.1 as the equal-weighted mean (in percentage) of monthly averages of $$Z_{t}^{\it FOR}$$, $$Z_{t}$$, and $$Z_{t}^{FX}$$, the daily fractions of ADRs in $${\it ADRP}_{m}$$ whose underlying foreign stock, ADR, or exchange rate experiences a zero return on day $$t$$. For each grouping, I list the ensuing total number of usable ADRs ($$N_{u}$$), as well as available months ($$N$$), mean, and standard deviation for each measure over the sample period 1980–2009. The“Other” grouping includes Colombia (COP), Denmark (DKK), Egypt (EGP), Hungary (HUF), Israel (ILS), New Zealand (NZD), Norway (NOK), Philippines (PHP), Singapore (SGD), Sweden (SEK), Taiwan (TWD), Thailand (THB), and Venezuela (VEF). Daily closing prices for these U.S. cross-listings, $$P_{i,t}$$, and their underlying foreign stocks, $$P_{i,t}^{\it FOR}$$, are also from Datastream. The corresponding exchange rates in Equation (11), $$S_{\it t,USD/FOR}$$, are daily indicative spot mid-quotes, as observed at 12 p.m. Eastern Standard Time (EST), from Pacific Exchange Rate Service (Pacific) and Datastream. Although commonly used, the resulting dataset allows to measure the extent of LOP violations in the ADR market only imprecisely (see, e.g., Ince and Porter 2006; Xie 2009; Gagnon and Karolyi 2010; Pasquariello, Roush, and Vega 2014). For instance, the trading hours in many of the foreign stock and currency markets listed in Table 1 are partly overlapping or nonoverlapping with those in New York, yielding nonsynchronous closing prices. Individual ADRP violations often differ in scale, making cross-sectional comparisons problematic, and either persist or display discernible trends. Paired closing foreign stock, currency, or ADR prices may also be stale (e.g., reflecting sparse trading), incorrectly reported (e.g., due to inaccurate data entry or around delistings), partly unavailable, or sometimes altogether missing. Pasquariello, Roush, and Vega (2014) proposes two measures of the marketwide (i.e., aggregate), low-frequency extent of violations of the ADR parity of Equation (11) addressing these concerns. The first measure, $${\it ADRP}_{m}$$, is the monthly average of daily equal-weighted means of all available, filtered realizations of $${\it ADRP}_{i,t}$$ of Equation (12), that is, of daily mean absolute percentage ADRP violations. In particular, I conservatively remove from these averages any available $${\it ADRP}_{i,t}$$ deemed “too large” ($${\it ADRP}_{i,t}>1,000$$ bps) or stemming from “too extreme” ADR prices ($$P_{i,t}<\$5$$ or $$P_{i,t}>\$1,000$$). These requirements and the aforementioned data limitations reduce the number of usable ADRs to $$319$$ in total and roughly uniformly across most groupings in Table 1, except for Turkey ($$N_{u}=3$$), Indonesia ($$3$$), Hong Kong ($$39$$), and Canada ($$46$$).21 Yet, filtering and daily averaging across individual ADRs minimize the impact of any idiosyncratic parity violations (or lack thereof), for example, due to quoting errors, missing data, or other data issues in the sample. Monthly averaging further smooths any spurious daily variability in observed ADRP violations, for example, due to bi