تعامل بین تجارت خارجی و بازده سهام در حال ظهور: شواهدی از ترکیه

Using monthly foreign flows data on Istanbul Stock Exchange (ISE) and employing a structural VAR model, we analyze the interaction between foreigners' trading and emerging stock returns. In contrast to most of the available theory and repeated previous findings on other markets, foreign investors negative-feedback-trade with respect to past local returns in ISE, however only in rising markets and especially under macroeconomic instability. Net foreign flows forecast future market returns, but not individual stock returns. Price impacts are permanent, suggesting that foreigners' trading incorporates information. Overall, results reject previous conclusions that foreigners are uninformed positive feedback traders: rather, they are a heterogeneous group dominated by sophisticated investors able to rationally adjust their trading style in line with the market's prevailing characteristics.

Many emerging economies have been dependent on international portfolio capital inflows, sudden reversals of which have been associated with severe destabilizing effects. Hence, policy makers and researchers are interested in understanding the nature of those flows and their impact on domestic financial markets. One strand of this literature studies the interaction between foreign investors' trading and equity returns, as foreign investors as a group are believed to behave differently and to have a strong impact on emerging stock markets. Foreign flows data compiled at the destination market enable a reliable analysis of foreign investors' trading behavior, the impact and forecast ability of their trades on stock returns. Specifically, recent research using such data (e.g. Dahlquist and Robertsson, 2004, Dvořák, 2005, Griffin et al., 2004 and Richards, 2005) has addressed the following questions: i) Do foreign investors pursue positive feedback trading strategies? ii) What is the impact of foreign flows on domestic stock returns? Is the contemporaneous impact explained by price pressure or by information? iii) Does foreigners' trading contain superior information (i.e., forecast ability)? Empirical answers to these questions, in turn, provide tests of theories of foreign investor behavior (Albuquerque et al., 2007, Brennan and Cao, 1997, Dvořák, 2003, Griffin et al., 2004 and Hau and Rey, 2004). EEMENA (Eastern Europe, Middle East, North Africa) region has been surprisingly neglected in this literature,1 despite it hosts those emerging economies that are most dependent on foreign capital inflows. The current empirical characterization of the interaction between foreign flows and emerging stock returns in the extant literature is incomplete as it excludes emerging economies with large external deficits in the EEMENA region. These markets differ from Asian markets extensively studied in the literature in that foreign investor participation is much higher: for example, the ratio of market capitalization held by foreigners has ranged between 60% and 75% in Turkey (as in many other EEMENA markets) in recent years, whereas the same ratio stands between 25 and 35% in Asian markets. While several panel studies include emerging markets from the EEMENA region, they use data either from one source country or one custodian.2 Such data have been shown to contain measurement errors as they do not include all foreign flows (Pavabutr and Yan, 2007). Indeed, the correlation between our actual and complete foreign flows series on Turkey and that derived from the Treasury International Capital (TIC) dataset is merely + 14.9% during our sample period. An exclusive branch of the literature has been emerging using complete data compiled at the destination market (Dahlquist and Robertsson, 2004, Griffin et al., 2004, Kim et al., 2008, Reis et al., 2010, Richards, 2005 and Samarakoon, 2009). The current paper contributes to this literature by presenting the first major study on an EEMENA market using complete foreign flows data compiled at the destination, both at the marketwide level and for individual stocks. The empirical evidence presented in this study offers an illuminating out-of-sample test of the aforementioned theories of international investor behavior. Istanbul Stock Exchange (ISE), the largest and deepest stock market in the EEMENA region, is an interesting avenue to add to this literature, as detailed foreign flows data are compiled in a centralized manner. Our comprehensive search process indicated that ISE offers one of the cleanest foreign flows data in the EEMENA region and the only publicly available data on foreigners' trading individual stocks.3 ISE is ranked seventh among all world emerging markets in terms of the total value of shares traded. Moreover, Turkish markets possessed some interesting characteristics such as persistent high inflation, very high real interest rates, political turnover and volatility during the first half of our sample period. A dramatic improvement in political and macroeconomic stability in the second half enables a comparison of foreign flows dynamics under different conditions (see Panel B of Table 1). Turkey removed all restrictions on foreign portfolio investments in August 1989. Available data on foreign flows starts from January 1997, therefore our analysis is not blurred by the initial impact of liberalization (i.e.; one-time portfolio rebalancing by international investors) as documented in Bekaert et al. (2002) and Dahlquist and Robertsson (2004). Finally, partial restrictions (e.g. Korea, Taiwan, Thailand) could distort the results in previous studies. Since 1989, Turkey has never implemented any (partial) restrictions on foreigners' trading in the stock market. Therefore, ISE offers an opportunity to document the true behavior of foreign investors free of any restrictions and post-liberalization effects, thus this study provides a major out-of-sample test of previous findings. Table 1. Panel A shows summary statistics. F is normalized net flows, R is local returns, W is global returns. Panel B shows selected indicators to compare our first and second subsamples. GDP differential is the average difference of Turkey's quarterly real GDP growth rate from the OECD average. Panel A F R W Mean 0.00049 0.0020 0.0037 St.Dev. 0.00278 0.1410 0.0459 ADF-stat. − 5.36⁎ − 13.67⁎ − 10.81⁎ Panel B 1997:1–2003:6 2003:7–2011-1 St.Dev. of real ISE returns 0.1821 0.0924 Average annual inflation 62.68% 9.97% GDP differential − 0.31% 0.91% ⁎ Implies statistical significance at the 1% level. Table options Employing one of the longest (14 years) samples in this line of literature, we present a comprehensive exploration of the interaction between foreign flows and equity returns, including several illuminating breakdowns missing in previous work. We compare and contrast the market-wide analysis with an analysis of foreigners' trading in individual stocks, which enables us to distinguish between alternative explanations in relation to the predictions theories of foreign investor behavior, as we shall discuss below. The bilateral interaction between net foreign flows and stock returns is analyzed within a structural vector autoregression (SVAR) model, augmented with world returns that are set to be exogenous to local variables. We also extend the VAR approach to individual stocks by using returns and net flows defined in relative terms. Our method offers a better alternative to the net buy difference method employed so far in this branch of the literature, as it enables us to measure the cross-sectional predictive content by identifying the surprise component of net foreign flows in individual stocks while the net buy difference methodology cannot distinguish the expected and surprise components. This approach can be utilized as an efficient procedure to distinguish informed trading by any investor group in individual stocks from market-wide informed trading. Such a distinction is important in this context, as foreign investors are believed to lack local information, yet their macroeconomic analysis skills may grant them market-wide forecast ability. The results provide detailed evidence on the feedback trading behavior and forecast ability of foreign investors. At the monthly frequency and the market-wide level, foreigners pursue positive feedback trading with respect to global returns, but negative feedback trading with respect to local returns. Net foreign flows have the ability to forecast future local returns over a horizon of several months. Our finding of negative feedback trading with respect to local returns contrasts most of the previous studies that almost uniformly report positive feedback trading by foreigners, and most of the available theory that describe foreigners as “uninformed positive feedback traders”. We explore the nature of the feedback trading behavior and the source of the forecast ability via a comprehensive set of break-downs. Further, by comparing and contrasting the market-wide analysis with individual stocks, we assess alternative explanations as to what drives negative feedback trading and investigate whether foreigners have a local information disadvantage. Specifically, if the negative feedback trading is motivated by the disposition effect or belief in mean-reversion, it should be present in individual stocks as well as market-wide. If, however, it is driven by cross-country portfolio rebalancing to manage exchange rate exposure, as argued by Hau and Rey (2004), it should be present only at the market-wide level, not in individual stocks. Based on a similar comparison, we can judge whether foreigners possess local information: if the forecast ability is confined to market returns, we can attribute it to their ability to analyze macroeconomic and global factors and conclude that they lack local information. This detailed analysis leads us to the conclusion that foreign investors are a heterogeneous group dominated by sophisticated traders who are able to rationally adjust their trading style in line with the market's prevailing characteristics. The paper is organized as follows: Section 2 reviews the literature addressing the research questions mentioned above, together with their theoretical backgrounds. Section 3 describes the data and methodology employed in this study. Section 4 presents results with illuminating break-downs to distinguish between alternative explanations, and Section 5 summarizes conclusions.

As previous empirical work on the joint dynamics of foreign flows and stock market returns surprisingly neglects the EEMENA region where economies most dependent on foreign capital inflows are located, the results of this study on ISE, the largest market in the EEMENA region, provides a major out-of-sample test of previous empirical findings and available theory. Furthermore, via a comprehensive set of break-downs, this study illuminates the nature of foreigners' feedback trading and the forecast ability of net foreign flows. Finally, this paper offers a unique study of foreigners' trading in individual stocks for the EEMENA region, as detailed data are available only for ISE. Extant research described foreign investors as “uninformed positive feedback traders”, which has been used as a justification for the argument that foreign portfolio flows may destabilize emerging markets given their size. Our findings indicate that foreigners negatively respond to past returns in ISE22 and that net foreign flows bear short-horizon forecast ability for future local market returns. This is in stark contrast with previous empirical results and most of the available theory that previously seemed consistent with empirical evidence. We then illuminate the nature of this feedback trading behavior via a set of breakdowns that enable us to assess alternative explanations. To begin with, the negative sign of the feedback rules out sentiment trading. Further, foreigners' negative feedback trading is confined to rising markets, which rules out naïve feedback strategies. The degree of negative feedback trading significantly decreased in the second half of our sample period, when a significant improvement in economic and political stability was achieved and the domestic individual investor participation significantly diminished. Interestingly, the forecast ability of market-wide net foreign flows is significantly reduced in the second half, too, which suggests that foreigners' contrarian trading with respect to local returns did counteract excessive bullish sentiment among domestic investors in a fragile and unstable economic environment in the first half.23 This can be interpreted to indicate that foreigners are sophisticated investors who can rationally adjust their trading style in line with the prevailing pattern of the fundamentals and the behavior of other participants, rather than naively pursuing a specific feedback trading or rebalancing strategy. In this respect, our results do not support theories based on foreign investors' informational disadvantages (Brennan and Cao, 1997 and Griffin et al., 2004), notwithstanding the possibility of intra-month positive feedback trading. Foreigners' purchases and sales are both positively related to local market returns contemporaneously, which is consistent with Albuquerque et al.'s (2007) result that within-country heterogeneity is more important than cross-country heterogeneity. However, upon a positive local return shock, the contemporaneous increase of foreigners' purchases outweighs that of sales, while in the subsequent month the increase in their sales outweighs purchases. Thus, while it is not adequate to treat foreigners as a homogenous group, it is a group clearly dominated by sophisticated investors who respond to information as it arrives and take profit after it is priced-in. The results on individual stocks, on the other hand, suggest no evidence of forecast ability, but the contemporaneous price impacts are permanent. Thus, although foreigners appear not to have any particular advantage with respect to local information, their trading incorporates information. We find little evidence of negative feedback trading by foreigners in individual stocks, which weakens the role of disposition effect as an explanation for (asymmetric) negative feedback trading. While this finding may be consistent with Hau and Rey's (2004) view that market-wide negative feedback trading is motivated by managing currency exposure, detailed break-downs point to more negative feedback trading during the first subsample and following positive abnormal stock returns, which may be a rational response to excessive instability. Hence, additional factors, not modeled in previous theories, may be playing a role in shaping foreigners' trading pattern: the stability of fundamentals and the degree of noise trading by local market participants. The same foreigners exhibit positive feedback trading with respect to world returns, which can be attributed to portfolio rebalancing (Griffin et al., 2004 and Kodres and Pritsker, 2002) as well as to a rational response to global macroeconomic information. Foreigners' prolonged response to world returns depicted in Fig. 7 supports our argument that monthly frequency is appropriate to measure the feedback trading behavior of foreign investors, since foreigners display a sluggish response to even world returns, which cannot be attributed to their informational disadvantages. We observe persistence in net foreign flows, consistent with Albuquerque et al. 's (2007) model prediction, which follows from heterogeneity among foreign investors. The degree of the persistence, however, is less than that reported in other studies, possibly due to the absence of post-liberalization effects. Net foreign flow persistence is weaker in individual stocks. The positive contemporaneous relation between net foreign flows and local returns is intriguing as it implies that domestic investors' trading is negatively related to current returns. As no subsequent reversals are observed, the contemporaneous price impact of foreign flows may reflect information and/or base broadening rather than price pressure. The results presented in this study collectively enable us to assess existing theories of foreign investor behavior. The model of Brennan and Cao (1997) attributes the positive contemporaneous correlation between foreigners' trading and local returns to local investors' cumulative information advantage, and Dvořák's (2003) model predicts a negative contemporaneous correlation due to local investors' marginal information advantage. Our results reject the latter outright. Negative feedback trading is inconsistent with models, including Griffin et al. (2004) and Albuquerque et al. (2007), that predict positive feedback trading due to an informational disadvantage. Yet, our analysis at the monthly frequency cannot rule out intra-month positive feedback trading, hence an informational disadvantage in terms of days. Finally, the market-wide forecast ability of net foreign flows is at odds with the widely-recognized model of Brennan and Cao (1997). If one considers that local returns may drive local investor sentiment while world returns is an unpredictable factor for local investors, it is possible to explain the local investor behavior documented in this study with a combination of sentiment and contrarian trading. Specifically, local investors appear to provide liquidity to foreigners responding to global information, and their bullish sentiment after local market increases offers foreigners a profit-taking opportunity. This is not only inconsistent with Brennan and Cao's (1997) assumption of cumulative information advantage of local investors, but also indicative of lack of sophistication. The observation that the degree of foreigners' negative feedback trading is substantially reduced in the second subperiod when local individual investor participation has decreased particularly fits this explanation. In sum, we show that previous findings that foreign investors are uninformed positive feedback traders may be premature. Rather, they are a heterogeneous group dominated by sophisticated investors who are able to rationally adjust their trading style according to market conditions and the amount of sentiment trading by local participants. They do not exhibit symptoms of uninformedness, which are underlying assumptions of models of international investor behavior. Rather, their response to local information is completed within the contemporaneous month, and in the following months they focus on rebalancing away from the host market. This raises serious doubt about the previously widespread stereo-typing of foreigners as uninformed positive feedback traders and the justifiability of policy suggestions to restrict their trading (the so-called “smart restrictions”).24 While similar conclusions have already been reached by other researchers over the last decade (e.g. Choe et al., 1999), renewed calls to reconsider such restrictions have been put forward following the 2008 global crisis. Since the current study is so far the only major analysis of this type from the EEMENA region which, in contrast to many Asian markets or Sweden, are characterized by large external deficits financed by foreign capital, an intriguing question is left for future research: Is foreigners' negative feedback trading driven by unique characteristics of Turkey such as high degree of instability and local investors being vulnerable to bullish sentiment, or common to all high-external-deficit economies? Available information suggests that foreign investor composition and characteristics in Turkey are no different from those in other emerging markets studied in the literature. Hence, our finding of negative feedback trading is not likely to be explained by unique characteristics of foreigners in ISE. On the other hand, large-external-deficit economies in the EEMENA region sharply differ from Asian economies in that whenever GDP growth accelerates current account deficits explode inducing concerns on the sustainability of growth, which makes negative feedback trading following local market increases rational. Our results indicated that the degree of negative feedback trading varies with the degree of instability in the host economy and local investors' vulnerability to sentiment. Obviously, a general answer to this question requires more evidence from other EEMENA markets, and the current paper opens up a new quest for future research.