2 Private information, stock markets, and exchange rates BIS working paper No. 271 February 2009 joint with M. Loretan, J. Gyntelberg and E. Chan of the BIS Tientip Subhanij 24 April 2009 Bank of Thailand Motivation 3 Related literature 4 Recent exchange rate literature has demonstrated that exchange rates are determined importantly by private information Macroeconomic data and other public information play a minor role once order flow is incorporated in the model The difference between buy and sell quantities initiated by a specific group of market participants An important question: do all categories of order flow matter equally for exchange rate determination? Related question: why do some types of order flow matter more? Early portfolio balance models and monetary models were found to perform poorly out of sample (Meese and Rogoff, 1983; Cheung, Chinn and Garcia Pascual, 2005) Subsequent research explicitly incorporated institutional features, i.e. the microstructure of asset markets and information held by traders. Insights from this research: traded quantities and prices reflect both institutional constraints and heterogeneity of information held by market participants. Order flow aggregates and conveys investors private information that is revealed only later in aggregate economic statistics (Evans and Lyons, 2007)
5 6 Intuition I Intuition II FX market makers observe their own customers order flow, as How to test this proposition empirically? well as public information. Should FX dealers change quoted prices in response to observed order flow? Yes, if order flow reveals information that is not already contained in the public information signal. Unfortunately, no FX flow data are directly tagged with flags for public or private information Indirect approach: find out if certain capital market flows are consistent with being driven mostly by public or private information. FX market makers must infer which part of their customers flows is associated with private information FX dealers optimal strategy is to change their bid and ask quotes permanently if the order flow comes from informed market participants (Bacchetta and van Wincoop, 2006). Our econometric strategy: Make use of the close linkages between the activities of foreign investors in the Thai equity and bond markets and their activities in the FX market. Hypothesize and confirm empirically that equity market order flow is driven mainly by private information, whereas bond market order flow is driven (exclusively) by public information Different reactions in response to 3 scenarios 7 or to put this into picture 8 Case 1 Public information T=0: A positive firm specific news item hits the stock market T=1: Because the signal is public, no trading (or very little trading) is necessary to achieve the new equilibrium. The effect of trades generated by public signal on the exchange rate is ambiguous (and likely small). Case 2: A pure liquidity trade T=0: A foreign investor sells a government bond, and sells the baht proceeds in the FX market to satisfy government regulations on permissible bank balances. All market participants understand that this is a pure liquidity trade. T=1: The price impact of this trade on the baht is minimal Case 3: Private information T = 0: A private information signal is picked up by local investors. They start bidding up the share price. T = 1: Foreign investors begin to pick up the private signal and start buying shares. Foreign investors acquire baht in the FX market. Baht appreciates as FX dealers react to customers order flow. Public information e.g. US macro data Liquidity trades e.g. mutual fund outflows Private Information e.g. Quality of firms products Order Flow Temporary/ permanent pressures on THB 8
The role of private information 9 Three empirical hypotheses to be tested 10 There are various types of private information that stock market investors may have, e.g. quality of firm s products, prospects for successful innovation, management quality, etc. 1. FX order flow of NR investors that is driven by their Thai stock market operations conveys private information and should have a large contemporaneous impact on the Thai baht. Equity market investors are generally very active in collecting private information, compared to other asset classes. Differences in private information between residents and NR 2. FX order flow that is driven by NR investors Thai government bond market operations does not convey private information and should have at most a small contemporaneous impact on the baht. investors will generate trades. These investors order flow conveys their private information to other investors and generates spillovers to other financial markets such as the FX market. 3. FX order flow that is driven by NR investors' Thai equity market transactions should also have a permanent effect on the exchange rate. Other portions of FX order flow should have at most a transitory effect on the exchange rate. FX order flow dataset 11 Basic aspects of our FX dataset 12 Non-resident By Contract Spot t Sample period: 3 Jan 2005-15 Dec 2006 The sampling frequency for all data is daily FX dealing Banks Sell FX Buy FX Resident Other banks Spot t+1 Spot t+2 Forward Swap All transactions flow data are for nonresident customers The flow data are constructed by aggregating across the daily Bank of Thailand By Purpose reports FX dealing banks must submit to the Bank of Thailand 12
Stock and bond market order flow 13 Data and description 14 Stocks (SET) Daily gross buy and gross sell transactions volumes by nonresident investors in Thailand Transactions on the SET settle on a T+3 basis Bonds (ThaiBMA) Daily buy and sell transactions volumes by nonresident investors in the government bond market on T+2 basis 13 Empirical framework I: Tests of hypotheses 1 and 2 15 A basic order flow regression 16 FX transactions are not tagged directly as being associated with stock or bond market transactions (or neither). Estimate portions of order flow that are related to stock and bond market variables using econometric techniques. This is possible because we have complete daily data for net transactions (order flow) in the FX, stock, and bond markets. Examine the relationships between the FX order flow proxies and exchange rate. 15 16
The big picture 17 18 Spot TD Components of FX order flow Spot TM Spot next Swap Forward FX Return Which part of order flows matter more for exchange rate? R 2 = 0.41 19 Empirical regularities I: Where is private information? Emp. regularities II: FX vs domestic capital markets 20 Order flow patterns of NRs in the FX and stock markets exhibit mild short-term flow momentum and return chasing. This is consistent with these transactions being driven at least in part by private information. Order flow of NR investors in the bond market does not exhibit flow momentum or return chasing in the sample, and therefore appears to be driven entirely by public information. We find that NR investors order flow in the two-day FX spot segment is highly correlated with their order flow in the stock market, but not with their order flow in the bond market. NR investors order flow in the one-day FX spot segment is highly correlated with their order flow in the bond market, but NR investors order flow in Thai stock and bond markets are nearly uncorrelated at the daily trading frequency. This implies that their stock market order flow is driven mainly by private information. not with their order flow in the stock market.
Empirical methodology 21 22 Step I Regress two-day FX spot order flow on stock market variables, retrieve fitted and residual series. (R 2 0.19) Regress one-day FX spot order flow on bond market variables, retrieve both fitted and residual series. (R 2 0.11) Step 2 Run an order flow regression in which fitted and residual series are both used as regressors. Use 2SLS procedure to adjust for fact that the FX order flow series are generated regressors. 22 Equity-market driven FX order flow have bigger impact on FX return 23 Summary of findings I 24 FX order flow Spot TD Spot TM Bond Others Spot next Stock Others Swap Forward FX Return R 2 = 0.48 Hypothesis 1 is supported by the data: Even though the coefficients of both the fitted and the residual spot-next series are significant, the coefficient of the fitted spot-next regressor is more than 3 times as large in absolute value as the coefficient of the residual spot-next regressor Inference: the portion of spot-next FX order flow that is driven by NR investors equity market activity has a more pronounced contemporaneous effect on the exchange value of the baht Hypothesis 2 is also supported: The portion of spot-tomorrow order flow that is related to NR investors transactions in the bond market does not have a significant impact on returns.
25 Further analysis 26 Which component of spot-next FX order flow has a permanent effect? In the preceding order flow regression, we found that both components of the spot-next order flow series -- the part fitted to spot market variables and the residual part -- are statistically significant in explaining contemporaneous returns to the baht Question: Does one of these two constructed order flow series have a more permanent influence on the baht than the other? Conduct a 3-variable vector autoregression (VAR) analysis with fitted part of spot-next order flow (i.e., fitted to equity variables), residual part of spot-next order flow, and THB/USD returns Calculate responses of THB//USD returns to innovations in each of the two order flow series 26 Effect of portfolio flows on exchange rate I 27 Effect of portfolio flows on exchange rate II 28 Equity-related spot-next flows Equity-related spot-next flows Other spot-next flows Other spot-next flows 27 28
29 30 Summary of findings II Alternative examination: Cointegration analysis Hypothesis 3 is supported: A one-s.d. innovation in the portion of spot-next order flow that can be attributed to equity market variables has an initial impact on baht returns that is almost twice as large as that of an innovation in the remainder portion of spot-next order flow. Initial impact of an innovation in spot-next FX order flow that is driven by equity market is not reversed, i.e. it has a permanent effect on exchange rate In contrast, the initial impact of an innovation in the residual portion of spot-next FX order flow is quickly undone, i.e. has A different way of examining which components of FX order flow have a permanent influence has been proposed by Killeen, Lyons, and Moore (2006): If FX order flow, an I(0) variable, has a permanent effect on the level of the exchange rate, an I(1) variable, then cumulative FX order flow and the exchange rate must be cointegrated. Conversely, if the (log) exchange rate and a cumulative order flow series are not cointegrated, the effect of order flow must be transitory. only a temporary influence on exchange rate 31 32 Hypothesis (H 0 ): Series has a unit root Summary of findings III (a) Univariate unit root tests LN(THB/USD) CUMUL(OF_SPOT_TOM) Reject Accept Four order flow series are I(1) while the other two series are I(0) We apply the test for cointegration between the I(1) order flow series and the (log) level of exchange rate CUMUL(OF_SPOT_NXT) CUMUL(OF_SPOT_TOM_FIT_SET) CUMUL(OF_SPOT_TOM_RES_SET) CUMUL(OF_SPOT_NXT_FIT_SET) CUMUL(OF_SPOT_NXT_RES_SET) Finding: The portion of cumulative spot-next order flow that is explained by stock market variables is cointegrated with the exchange rate Cumulative total spot-tom order flow, as well as the portion that (b) Unit root tests on residuals is fitted to the stock market variables, also cointegrated with LN(THB/USD) vs. CUMUL(OF_SPOT_NXT) LN(THB/USD) vs. CUMUL(OF_SPOT_NXT_FIT_SET) LN(THB/USD) vs. CUMUL(OF_SPOT_TOM) LN(THB/USD) vs. CUMUL(OF_SPOT_TOM_FIT_SET) exchange rate We conclude that the portion of FX order flow that is explained by stock market variables has permanent effects on baht
Concluding remarks 33 34 We confirm previous solid findings in the FX determination literature on the importance of order flow. Main contribution of this paper: We add to the scholarly literature by showing that the order flow relevant for exchange rates is related to the stock market, and that this relationship holds because this order flow conveys private information. We find that even though FX flows that convey private information amount to only a very small part of all FX flows, this subset of flows is very influential. Further work on identifying those flows that convey private information would be useful It would also be useful to investigate the role of other types of order flows (banks and residents, inter-bank o/f) in determining exchange rate movements Future work may also include an analysis of the information content of intra-day order flow data for exchange rate determination 33 Thank you