Master’s Thesis
Market Efficiency and Stock Return Predictability:
Emerging countries vs. Developed countries
Student name: Bogdan Marinel Student number: 10621091
Date of the final version: July 1, 2018 Programme Code: MSc FIN
Specialization: Asset Management Supervisor: Esther Eiling
Statement of Originality
This document is written by Bogdan Marinel who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
Stock return predictability is a heavily studied topic in asset pricing. Researchers are trying to predict excess stock returns using financial, economic, and business cycle variables for decades. On top of the previously studied predictors, a new predictor is proposed: exchange rate returns. Past literature evidences that stock markets in emerging countries incline towards inefficiency while exchange rate markets are leaning towards efficiency. We test if the more efficient market returns are able to predict the inefficient market returns. This paper examines the predictive power of exchange rate returns in both emerging and developed countries, compares the results between countries and aims to shed light on the role of efficiency in stock return predictability. The overall results suggest that emerging countries excess stock returns are predictable by exchange rate returns. However, as we also find predictability in developed countries stock returns, a final conclusion cannot be drawn with respect to market efficiency.
Contents
1.
Introduction ... 1
2.
Literature Review ... 4
2.1. Emerging Stock Market Efficiency ... 4
2.2. Developed stock market efficiency ... 6
2.3. Emerging exchange rates market efficiency ... 7
2.4. Developed exchange rates market efficiency ... 8
2.5. The relationship between exchange rate and stock market returns ... 9
2.6. Past studies on predictability... 11
3.
Data & Methodology ... 13
3.1. Data ... 13
3.2. Methodology ... 16
4.
Preliminary Results ... 19
4.1. In-sample analysis using USD/domestic exchange rate returns as predictor ... 19
4.2. Out-of-Sample analysis using USD/domestic exchange rate returns as predictor ... 25
4.3. Inflation rate as an alternative predictor... 30
4.4. In-sample analysis using EUR/domestic exchange rate returns as predictor ... 34
4.5. Out-of-sample analysis using EUR/domestic exchange rate returns as predictor... 38
4.6. Exchange rate predictability from an US investor standpoint... 43
5.
Conclusion & Future Research ... 49
Bibliography ... 51
1
1. Introduction
Emerging economies have grown rapidly over the last years, with China and India accounting for 40% of the growth, becoming a worthwhile choice for investors. However, even if China and India stole the spotlight, there are other emerging markets less visible for the public that could offer new and profitable investment opportunities. According to Harvey (1995), emerging countries markets offer high expected returns but also high levels of volatility to investors. The most important feature of emerging markets equity is the low correlation with developed markets equities, meaning that an investor that invests in both emerging and developed countries can significantly decrease its portfolio risk.
One of the core research areas in asset pricing is how the equity premium varies over time. Asset pricing researchers spent the last two decades exploring whether stock returns can be predicted by a vast range of factors. The efficiency level of a stock market may be an important characteristic in determining its predictability. As efficiency is related to how fast new information is priced in the stock value, one may argue that if the stock market returns are not predictable then the market is efficient. On the other hand, if a market is predictable, its efficiency level is questionable. Mutually, efficient and inefficient markets may show signs of predictability, sometimes arising due to time-varying risk. As emerging countries have less developed and transparent stock markets than developed countries, their equity market efficiency level is uncertain.
Recent studies suggest that emerging countries equity markets are weak-form efficient or inefficient while emerging currency markets are inclining towards efficient. It comes rather unexpected that emerging countries currency markets are more efficient than stock markets. This, however, is a curious matter. As one of the markets has most available information already priced in while the other market does not, we expect the efficient market to be able to predict the inefficient one if both markets are driven by the same fundamentals. The difference in the level of efficiency between the two market types can fill in a literature gap and support further stock returns predictability research: if using the efficient market returns to predict the weak-form/inefficient market returns within the same country indeed improves excess stock returns predictability. On the other hand, past literature confirms that developed countries stock markets incline towards efficiency while their currency markets tend to be less efficient. Thus, to be able to conclude that market efficiency is the key component of
2
predictability in emerging countries stock returns, we expect to find no predictability in developed countries stock returns. There are two questions this paper aims to answer: How predictable are stock returns in emerging countries by their currency returns? Is the effect present in developed countries?
To answer our research questions we perform a time-series stock return predictability methodology on a time-period of 30 years: 01/04/1988-01/04/2018. We run regressions both in-sample and out-of-sample. The analysis is conducted on 20 emerging countries, selected based on data availability, and 5 developed countries, selected based on the importance of the country (G-7 countries) but also accounting for their distinctive domestic currencies. Firstly, we test for stock return predictability in both emerging and developed countries using the exchange rate USD/domestic. Secondly, we compare the results to an alternative predictor, the inflation rate. To be able to prove that our results are recurring we perform the same analysis with EUR/domestic exchange rate returns as predictors. Lastly, our robustness check tests for predictability from an US investor point of view, converting all excess stock returns to US dollar returns and testing for predictability using the USD/domestic exchange rate returns. We conclude by comparing predictability between emerging and developed countries excess stock returns.
Our main findings are as follows. Emerging countries excess stock returns are predictable both in-sample and out-of-sample using USD/domestic exchange rate returns. 16 out of the 20 emerging countries excess stock returns are predictable in-sample. At K=1, 6, 12 most emerging countries have negative relations between their currency returns and excess stock returns. On longer horizons, the signs shift, positive currency returns predicting higher excess stock returns. One explanation may be that, when currency returns are positive (domestic currency devaluates), investors perceive the domestic country as “riskier” and avoid investing in the selected country, leading to lower excess stock returns in the short-run. However, because the country’s currency becomes cheaper, its competitiveness on the export market increases, leading to long-run improved excess stock returns. 14 out of 20 emerging stock market returns are predictable out-of-sample. There are weak signs of out-of-sample predictability on short horizons; however, predictability strengthens on longer horizons. Developed countries excess stock returns show signs of predictability as well when USD/domestic exchange rate returns are used. We find in-sample predictability in 3 out of 5 developed countries excess stock returns. In-sample predictability arises in up to 12-months ahead excess stock returns. Across all horizons, developed countries currency returns show a
3
positive relation with excess stock returns. Out-of-sample predictability is found in 4 out of 5 developed countries excess stock returns. The highest values of developed countries OOS are found when 12-months ahead excess stock returns are predicted. Further, we compare the exchange rate returns predictive power to the one of the inflation rate, finding similar results. As Argentina Consumer Price Index (CPI) data is not available, we exclude Argentina from the sample when running the alternate predictor regressions. We find in-sample predictability in 16 out of 19 emerging countries excess stock returns and out-of-sample predictability in 15 out of 19 when we use the inflation rate as predictor. In-out-of-sample predictability arises in 4 out of 5 developed countries and out-of-sample predictability arises in 3 out of 5 countries. The 3rd step of our analysis encompasses EUR/domestic exchange rate returns predictive power. Both in-sample and out-of-sample, excess stock returns are predictable in 17 out of 20 emerging countries. On the other hand, developed excess stock market returns are predictable in-sample in only 1 country and out-of-sample in 4 out of 5 countries. In the last step of the analysis, the robustness of the results is tested by studying the predictive power of USD/domestic exchange rate returns from an US investor perspective. In-sample, we find less emerging countries with predictable excess stock returns. The emerging countries with significant relations between their currency returns and excess stock returns show similarities in the nature of their relations when compared to the results found when predictability is tested from a domestic investor point of view. In the short-run, higher currency returns predict lower excess stock returns, however in the long-run the results are mixed. Though the in-sample results tend to differ, we find out-of-sample predictability in 12 out of 20 emerging countries and 4 out of 5 developed countries, allowing us to conclude that the initial results are robust.
Generally, emerging countries excess stock market returns prove to be predictable both in-sample and out-of-sample. They are indeed more predictable by EUR/domestic exchange rate returns than by USD/domestic ones, one explanation being the domestic macroeconomic influences on exchange rate fluctuations and the sovereignty of exchange rate guidance of the reference currency country’s central bank. On the other hand, we also find predictability in developed countries excess stock market returns. These findings tend to disregards our main intuition: that the difference in market efficiency levels plays a role in predictability. However, we cannot draw a concrete conclusion regarding the different levels of market efficiency role in predictability as overall predictability may also arise due to the macroeconomic implications of the exchange rates or other factors. To be able to further
4
research the implications of market efficiency on stock return predictability, we recommend using the market efficiency measure proposed by Kristoufek and Vosvrda (2013). Knowing each country’ level of market efficiency may clarify if the difference in levels of efficiency between currency and stock markets has an effect on predictability.
The development of the rest of the paper is conducted according to standard academic practice. At first, a theoretical framework is shown, providing past relevant researches and theoretical reasoning. Secondly, data and methodology is discussed; this section includes details on the selected observations and variables, variables computations and the approach of the research. Following, the analysis of the study results is described. Lastly, arguments are assessed and conclusions are drawn, outlining several limitations and proposing recommendations for further research.
2. Literature Review
One of the first academics that studied market efficiency is Fama (1970). He states that resource allocation should be done based on price signals and calls this “the ideal market”. An “ideal market” is one in which stock prices fully reflect all the available information of that market. If all available information is reflected then a market can be defined as “efficient”. Markets may experience three different efficiency levels, weak form efficiency, semi-strong form efficiency and strong form efficiency. A market is weak form efficient when its prices are adjusted to historic prices and only the historic prices deliver relevant information. Semi-strong form efficiency analyzes if other publicly available information beside historic prices is incorporated into prices. Lastly, the prices of the markets that are strong form efficient may be adjusted to private information that an individual or a group of investors have access to.
2.1.Emerging Stock Market Efficiency
There are several studies on stock market efficiency in emerging and developed countries. Lagoarde-Segot and Lucey (2006) experiment stock market efficiency in Middle-Eastern North African stock markets by running random walk tests and technical trade analysis and found that emerging Middle-Eastern North African stock markets are weak-form efficient. Hamid et al. (2017) found in a recent study that 14 emerging countries from Asia-Pacific region have inefficient stock markets, concluding that investors may get extra profits from
5
arbitrage strategies due to the inefficiency in the markets. In line with emerging countries stock markets efficiency is the research of Morck et al. (2000). The authors test the stock return synchronicity for both emerging and developed countries. Their findings suggest that stock markets in emerging economies are less useful at processing information and face more politically driven shifts than developed countries stock markets, thus stock prices in emerging countries tend to move together. One of the reasons behind the slow process of information in emerging countries and stock returns synchronicity is analyzed by Bhattacharya et al. (2000) in their study on Mexico. They test and find that the Mexican stock market does underreact or does not react at all to events on the announcement day. The main explanation of the close to zero effect on stock prices is inside trading. The author tests for inside trading by splitting investors into two groups, domestic investors and foreign investors and their reaction to events; the results show that foreign investors are more surprised than domestic investors and that the lack of reaction is present mainly for domestic investors.
Fama (1995) describes the random-walk model as an accurate description of reality that dismantles several other models and technical procedures. If a country’s stock market returns follow a random-walk behavior then the stock market is characterized by weak-form efficiency. Grieb and Reyes (1999) re-examine in their paper the random-walk behavior of Brazilian and Mexican stock market indices. They find that the Mexican stock market follows a mean aversion pattern, rejecting the random-walk behavior. On the other hand the Brazilian indices follow a greater tendency towards random walk, however when firms are individually analyzed, they follow a mean reversion behavior. Latin American stock market efficiency is the focus of Worthington and Higgs (2003) paper as well. They test for random walks in seven South American countries stock markets using serial correlation coefficients and runs test. The outcomes of the tests are that random walks are not present in any of the selected stock markets, meaning that the South American countries do not have weak-form efficient markets. Another study on the random walk behavior of emerging countries stock markets is the one of Abraham et al. (2002), focusing on three emerging Gulf countries: Saudi Arabia, Kuwait and Bahrain. Their methodology is slightly different than the ones of the previous papers: because the tests for market efficiency are imprecise due to infrequent trading in the selected countries, the authors correct the index returns for infrequent trading. Their results are in line with previously discussed papers on emerging countries; random-walk and weak-form efficiency is rejected. However, when the index values are controlled for infrequent trading both Bahrain and Saudi Arabia markets are following a random-walk, implying
weak-6
form efficiency, while Kuwait does not follow a random-walk even after controlling for infrequency.
2.2. Developed stock market efficiency
A comparison between emerging and developed countries stock market efficiency is done by Rizvi et al. (2014). The study is conducted on 21 countries stock market index: 11 developed countries and 10 Islamic countries (emerging countries) during the time-period 2001-2013. The authors’ test and rank countries by stock market efficiency levels, concluding that developed countries are relatively more efficient than the Islamic countries. Furthermore, developed countries stock markets are highly efficient in the short-run and moderate efficient in the long-run. The results of a relevant paper, entitled “Ranking efficiency for emerging countries” written by Cajueiro and Tabak (2004), are in line with the above findings. The countries analyzed are Japan, US and 11 emerging countries from Asia and Latin America. The authors evaluate stock market long-run dependency and rank stock markets by their efficiency levels. Their results suggest that Japan and US stock markets are leaders in market efficiency, followed by Latin American stock markets and lastly the Asian stock markets. Lim and Brooks (2005) conduct a similar study to the one of Cajueiro and Tabak; they capture the evolution of stock market efficiency of 23 developed countries and 27 emerging countries and compare the level of efficiency between them. Their results show higher levels of stock market efficiency in developed markets than in emerging markets. However, there are some exceptions that suggest that market efficiency does not depend on the level of development of the market, an example being Taiwan. Lim (2007) published a more recent paper, with the same objective as the one written with Brooks in 2005, but on a smaller country sample and with a higher focus on methodology. His standalone paper is highly influenced by the study of Cajueiro and Tabak, focusing on the same 11 emerging countries and 2 developed countries; nevertheless the contribution of the paper is given by the demonstration that nonlinear dependence in stock returns also evolve over time. The results show that the US stock market is the most efficient market while the Argentinian market suffers the most frequent drifts from efficiency. Another study on developed countries stock market efficiency is the one of Chan et al. (1997). The authors evaluate stock market efficiency and integration in 18 countries, 16 developed and 2 emerging countries. Their results suggest that all 18 countries stock market is at least weak-form efficient and that diversification among the selected countries may be effective as there are few signs of
7
cointegration between stock markets: long-run stock movements tending to differ between each country.
The difference between the efficiency levels of stock markets in emerging and developed countries is vital for the validity of this paper. With the confirmation from past researches that developed stock markets are more efficient than emerging stock markets, there is only one missing piece of the puzzle to completely support the reasoning of this study: the reverse efficiency level in the currency market. If developed stock markets are more efficient than emerging stock markets but developed currency markets are less efficient than emerging currency markets, then predictability might arise from the different levels of efficiency between the two types of markets: the more efficient currency market predicting the less efficient stock market, with the sole focus on the less efficient emerging stock markets predictability by its more efficient currency market.
2.3. Emerging exchange rates market efficiency
Evidence on currency market efficiency of emerging and developed countries is mixed. Hsu et al. (2016) found that when comparing the predictive power of currency markets in developed and emerging countries, emerging countries are more predictable with technical analysis, therefore if there is more predictability, markets may be efficient, but it could also be a sign of inefficiencies. However, the findings of Hsu et al. are supported by the paper of Sensoy and Tabak (2016); their outcome suggests that exchange rates in developed countries are less efficient than in emerging countries. Lima and Tabak (2007) test in their paper foreign-exchange rates market efficiency of eight emerging countries. Their results suggest that emerging countries that recently adopted floating exchange rate regimes improved on efficiency. Nevertheless, random walk behavior cannot be rejected for the countries sample, meaning that the foreign-exchange markets of the selected countries may be efficient. A study that supports foreign-exchange efficiency is the one of Wickremasinghe and Kim (2008) on Sri Lanka. Using different unit-root tests the authors test Sri Lanka rupee efficiency when paired with Indian rupee, UK pound, US dollar and Japanese yen. The outcome suggests that while the exchange rate is floating there is efficiency in Sri Lanka rupee when paired with all four currencies. Da Silva et al. (2007) evaluate the exchange market efficiency of Brazil and find that Brazilian daily foreign exchange market is weak form efficient. However, their findings show that the efficiency of the market was declining since the 1999 crisis. .
8
2.4. Developed exchange rates market efficiency
Hakio and Rush (1989) analyze the market efficiency of the sterling and deutschemark exchange markets. The authors test several aspects of market efficiency. Their overall outcome is mixed, however their strongest results suggest lack of efficiency for both markets. In line with the above effects are the results of Cheung (1993) in his study about long memory of currency market returns. If the market returns have long memory then the random-walk hypothesis is rejected, thus efficiency is rejected. Cheung results suggest that United States, United Kingdom, West Germany, Switzerland, Japan and France foreign-exchange markets do not follow a random-walk, following more complex dynamics than inferred by random walk and show signs of long memory. Souza et al. (2008) analyze long memory in European currencies within the European Monetary system (EMS). Their results suggest that currency markets of the countries within the EMS have long memory and that the long memory occurs mainly because of they are part of the EMS. They prove this fact by focusing on the United Kingdom. United Kingdom was part of the EMS, having long memory in their currency, highly amplified during the crisis. However, United Kingdom divided from the EMS after the European financial crisis. The outcome of the exit lead to higher efficiency levels in the currency of the United Kingdom. Another study on developed countries exchange markets is the one of Ahmed and Ansari (1997) on the exchange market of Canada. They use the transfer function analysis to test the efficient market hypothesis. The results suggest that in the selected sample period there were arbitrage opportunities, leading to the rejection of the efficient market hypothesis.
According to Burnside et al. (2010), carry trade is one of the oldest and most used trading strategies on the exchange rate market. Carry trade is a speculative strategy that consists of burrowing low-interest currencies and lending high-interest ones. Academics are not able to fully clarify the implications of this strategy as they are not able to explain its specious profitability. Brunnermeier et al (2008) state that exchange rates tend to unexpectedly shift even when related news are not released, one reason behind these movements being the carry trade investors hitting the funding constraints and reducing the number of carry trade deals. Heath et al. (2007) consider that carry trades are particularly hard to track, thus, country exchange rate market efficiency may be linked to the volume of ongoing carry trades.
Past literature is able to support our initial statement: the emerging countries stock market is inefficient while developed countries stock market leans towards efficiency and emerging countries exchange rate market is more efficient than the developed one. As exchange rate is
9
a macroeconomic variable before anything else, it is exposed to economic instabilities, reflecting the state of a country’s economy. Emerging countries exchange rates tend to be efficient, meaning that most information is priced in. On the other hand, emerging countries stock markets are inefficient, meaning that relevant “news” are not priced in on announcement. However, the information that is not priced in the stock prices now may be priced in at a delayed date, thus, as exchange rates values already reflect the delayed information, predictability may arise.
When it comes to developed countries exchange rates, they tend to be less sensible to information, inclining towards inefficiency. We believe that exchange fluctuations in emerging countries are the results of internal influences while developed countries are more reluctant to severe monetary policies. Let us consider the Eurozone countries as an example: Eurozone consists of 19 countries that use the same currency, significant exchange rate fluctuations are less likely to happen as such fluctuations may affect the 19 countries in 19 different ways, leading to increased levels of economic instability. Another example may be the impact of currency fluctuation on the world’s economy: a big swing in the US Dollar rates may affect the global economy as a whole. Hence, the lack of efficiency in developed countries exchange rates. As developed countries exchange rates contain no “new” information that may influence stock returns, it is expected to find no stock returns predictability by their exchange rates.
2.5. The relationship between exchange rate and stock market returns
The contemporaneous relation between exchange-rate movements and stock returns has been heavily studied. In Ma and Kao (1990) research the reaction of stock prices to exchange rate changes is examined. They study two different currency return effects on stock returns: the financial effect and the economic effect. The financial effect is related to the exposure that investors face when investing in a foreign currency, the exchange rate volatility. According to the authors, investors are attentive to investments that are denominated in strong currencies, creating a link between high exchange rate levels and advantageous stock price movements. The economic effect is related to firms’ competitiveness on the export and import markets: if domestic currency depreciates then the export firms increase in competitiveness, leading to a positive effect on the domestic stock market. On the other hand, for firms that are import oriented, a currency depreciation increases import costs leading to a
10
negative effect on stock prices. Overall the effect is ambiguous and can go both ways, depending on the economic development of each country.
A study on the intertemporal relation between stock indices and exchange rates is the one of Ajayi and Mougouė (1996). The authors test the short and long-term dynamics of the relation between the two variables of eight developed countries using time-series analysis. Cointegration is found between the two variables time-series for all countries. The results suggest that if stock prices increases then the currency devaluates in the short-run and appreciates in the long-run. Correspondingly, currency depreciation leads to a decrease in stock prices in the short-run but also in the long-run. Dominguez and Tesar (2006) examine the relation between firm value and exchange rate movement in 8 Non-US countries. A statistically significant relation has been found between profitability, measured as stock returns, and the exchange rate. In their paper, Inci and Lee (2011) study the effect of exchange rate movements on stock returns in 5 developed countries and found that lagged exchange rate movements impact stock returns. The same authors suggest that an investor that hedges against currency changes gains higher returns than an investor that does not hedge.
There are two important studies on the relationship between exchange rates and stock returns in emerging countries, the ones of Abdalla and Murinde (1997) and Walid et al. (2011). Abdalla and Murinde found strong evidence on the causal influence of exchange rates on stock prices in emerging countries. The paper of Walid et al. emphasizes on the role of exchange rate changes in emerging countries stock market: “exchange rate changes play a significant role in determining the switch between calmer and more turbulent periods in emerging stock markets” (2011, p. 290). Bailey and Chung (1995) conduct a study on the effect of exchange rate fluctuations and political risks on Mexican stock returns. Their study aims to find evidence on the impact of exchange rate fluctuations and political instability on the stock prices of the individual Mexican companies. The reasoning behind their hypothesis is that domestic firms suffer if exchange rates are volatile and hedging against them is costly (foreign competition, increased costs etc.). These effects may impact firms share price. They run a multifactoral model, with exchange rate and political risk as chosen factors, and study the cross-section implication of the stock returns when exposed to those factors: if the stock prices are driven by their factors. The conclusion of the paper is that the authors found some evidence regarding time-varying equity market premiums exposure to changes in the exchange rate market premium, resulting into a significant relation between stock market premiums and currency market premiums for Mexico.
11
A relevant paper is the one of Ajayi et al. (1998) in which the relationship between stock returns and exchange rates in seven developed countries and eight emerging Asian countries is studied. Their findings suggest that there are causal relations between stock indices and exchange rates in all developed countries and weak evidence of causality in emerging countries. The authors’ explanation is that the distinction of the results regarding developed and emerging countries is the development level of the stock market. This could indirectly imply the difference in stock market efficiency between developed and emerging countries when it comes to exchange rate “information”. While for developed countries the causal relation exists, the exchange rate volatility may already be priced in the stock market index: stock market can be efficient towards the exchange rate available information. On the other hand, the emerging countries equity market efficiency is questionable as the causal relation is not clear. This may lead to the stock market not pricing in the shocks of exchange rates and thus leading to stock market predictability.
The above literature empowers the intuition of our paper, establishing a link between stock returns and currency returns, by knowing that exchange rates, to some extent, play a role in determining stock returns.
2.6. Past studies on predictability
Several studies on the topic of stock return predictability concluded that future stock returns are predictable using publicly available information, such as macroeconomic variables that impact firms’ business cycle (Pesaran & Timmermann, 1995). The reasoning behind these studies is the indirect influence of the state of the country in which the firms are active on overall firm performance. Some factors that indirectly influence firms are inflation rate (example: Bodie, 1976), unemployment rate (Boyd, Jaganathan & Hu, 2001), interest rates (Ang & Bekaert, 2001) and so forth. Depending on the variable the outcome differs. Numerous studies found evidence about macroeconomic variables predictive power on stock returns. On the other hand there are studies on several macroeconomic variables in which the results suggest strong evidence on predictability while other studies on the same variable did not find any evidence, thus data mining concerns arise (Rapach et al., 2005).
Most stock returns predictability papers have as focus the United States stock market. Data availability may be one of the reasons behind this emphasis. Boyd et al. (2005) study the reaction of the US stock market to unemployment announcements. Their results suggest that stock prices rise when there is bad unemployment news during economic expansions and
12
decrease during economic contractions. There are two factors that may drive stock prices when compared to the government securities: equity risk premium and expected dividend growth rate. According to the same authors, unemployment news holds information about at least one of the two factors. Patelis (1997) tests stock return predictability using monetary policy changes. By using both long-horizon regressions and short-horizon regressions the author finds new variables that significantly predict stock returns. The variables used are the federal funds rate, the spread between the federal funds rate and the yield on the ten-year Treasury note, the quantity of non-borrowed reserves, the spread between the yield on six-month commercial paper and six-six-month T-Bills and the portion of non-borrowed reserve growth independent to total reserve growth. Both short-term and long-term predictability is found using the above variables. Another important paper on stock return predictability is the one of Ang and Bekaert (2007). They study the forecast power of dividend yield towards excess returns, cash flows and interest rates in United States, United Kingdom and Germany. The prediction power of dividend yield on future excess returns is mixed, meaning that it has prediction power on short-horizons however there is no prediction on longer horizon, the results are not robust across countries and across sample periods. On the other hand the authors find evidence on the prediction power of dividend yields on interest rates and cash flow; however these results are beyond the scope of our study.
Harvey (1995) analyzed stock return predictability in emerging markets and found that, compared to developed countries markets, emerging markets are more likely to be influenced by local information. As currency markets in emerging countries are inclining towards efficiency, meaning that all the available information is already priced in, one could argue that these returns are an accurate representation of the state of the economy and thus be a strong equity returns predictor. Another study on stock return predictability in emerging countries is the one of Narayan et al. (2014), who conducts an in-sample, out-of-sample and trading strategy methodology on the return predictability of emerging countries. Their findings prove that by using macroeconomic and institutional factors the returns are predictable in-sample and investors can make significant profits from dynamic trading strategies. In his paper he uses a similar methodology to ours but their analysis is conducted on different predictors and sample. Chang et al. (2004) test stock return predictability in eleven emerging countries using multivariate variance ratios. They first show that emerging countries do not follow a random walk, meaning that weak-form efficiency can be rejected. Secondly they use variable moving average and trading range break technical rules and prove that there is predictability in the returns. Lastly, they control for trading costs and
buy-and-13
hold strategies, leading to a change in their first predictability results: a decrease in overall predictability. Claessens et al. (1995) examine the behavior of twenty emerging countries stock markets. They test for return anomalies, with seasonality as their main focus, and predictability. The results show that most emerging countries suffer from the same anomalies, but they also find stock return predictability in industrial markets. However, the authors are not able to determine if predictability is caused by the lack of market efficiency or time-varying risk premiums, the conclusion being that the high level of predictability arises primarily from the lack of development and transparency of the emerging markets, threatening the uninformed investors.
We consider the methodology of Welch and Goyal (2007) to be one of the best approaches in achieving the purpose of this paper, as we are interested in the predictive power of an individual variable: the exchange rate returns with respect to excess stock returns.
The contribution of this paper is to evidence if exchange rate returns are a valid predictor of excess stock returns in emerging countries. Correspondingly, to verify if the different levels of market efficiency between stock and currency markets play a role in stock market predictability. It comes to our knowledge that this is the first paper that conducts such an elaborate study on exchange rate predictive power and its implications in stock return predictability.
3. Data & Methodology
3.1. Data
To answer the research question, the analysis is conducted on 20 emerging countries, selected based on data availability, along with 5 developed countries with which the emerging countries are compared. The selection of the developed countries is influenced by the importance of the countries (main focus is on the G-7) and also by the versatility towards domestic currency. Data is available on DataStream and is gathered on the 30-year time window 01/04/1988-01/04/2018. A large number of observations are required for a predictability analysis in order to find meaningful results. Not all emerging countries have data sets on such long time span due to their stock markets late development. Yet we conduct the same analysis on these countries, considering that this is not affecting the validity of our paper. Their outcomes should, however, be interpreted with more caution.
14
The data essential for this study contains: stock index excess returns, country exchange rate returns with respect to the EURO and the US Dollar, as well as a country specific risk-free rate proxy and inflation rate. Excess stock returns are our dependent variable while exchange rate returns, risk-free rate and inflation rate are our independent variables.
Our dependent variable is monthly excess stock returns of the selected countries main index. In order to construct them, monthly observations of the total return index price levels are gathered from DataStream. The market returns are computed individually for each country as follows:
where is the index return at time t, is the index price level at time t, is the index
price level at time t-1 and t has a monthly frequency.
To obtain the excess market returns, one should subtract the risk free rate of the country from the index returns. Risk free rate cannot be directly observed, however there are several proxies that measure it. As data availability is problematic for emerging countries, the proxies chosen among countries may differ, but always one of the following four variables: 3-month Treasury bill rate, 10 year government bond yield, interbank rate or deposit rate. Risk free rates are then computed as:
where is the risk free rate at time t, is the risk free rate proxy at time t and t has a monthly frequency.
This leads us to the last step of the monthly excess stock market returns computation:
where is the monthly excess stock market index return at time t, is the index return at time t, is the risk free rate at time t and t has a monthly frequency.
Further, the exchange rate returns are computed. Currency pairs USD/domestic and EUR/domestic monthly values are used. As Euro was introduced in 1999, exchange rate observations with respect to EURO on the time-period 1988-1999 are computed using DataStream synthetic Euro rate. The computation is similar to the one of index returns:
15
where is the currency return at time t, is the exchange rate at time t, is the
exchange rate at time t-1 and t has monthly frequency.
Intuitively when is positive it means that the domestic currency is depreciating, respectively appreciating when is decreasing. As an example, let’s consider Romania and take as reference the USD/RON rate. At time t the exchange rate is 3.81 and at time t-1 the exchange rate is 3.74. This implies a return of 1.85%, meaning that the domestic currency has depreciated by approximately 1.85% related to the US dollar.
To assess the robustness of the results, we perform the analysis on both the USD/domestic and EUR/domestic exchange rates, two reference currencies with completely different underlying economies. Predictability is also tested with and without control variables. As a last robustness check, the prediction power of the USD/domestic exchange rates is tested from an US investor viewpoint, meaning that countries stock market returns are transformed in US denominated returns and the US risk-free rate is subtracted from the returns in order to find countries US denominated excess stock returns. Thus, if predictability arises from both exchange rate returns with and without controls we can conclude that our results are robust.
The US denominated returns are constructed as follows:
( )
where is the US denominated stock return at time t, is the domestic denominated stock return at time t, is the currency rate change (USD/domestic) at time t and t has monthly frequency.
Further, US denominated excess returns are computed:
16
where is the US denominated excess stock return at time t, is the US denominated
stock return at time t, is the US risk-free rate according to Kenneth R. French website1 at time t and t has monthly frequency.
Our main control variable is the demeaned risk free rate. As risk-free rate is an important element of the excess stock market returns and also a volatile variable in emerging countries, controlling for it may lead to increased predictability when paired with the exchange rate returns. Thus, we control by using a zero-mean variable that represents the risk-free rate. This enables us to conclude if exchange rate returns predictive power is amplified when demeaned risk-free rate is added to the analysis. We compute the variable as follows:
∑
where is the demeaned risk free rate at time t, is the risk free rate proxy at time t,
∑
is the weighted-average of the past 12 risk free rates and t has monthly
frequency.
To be able to compare our results to the ones yield by already studied predictors, we compute excess stock return predictability by inflation rate. As inflation is an important macroeconomic factor, we expect similar results to exchange rate predictability. We compute inflation using Consumer Price Index (CPI) values:
where is the inflation rate at time t, is the inflation price index at time t, is the inflation price index at time t-1 and t has monthly frequency.
3.2. Methodology
We adopt a time-series stock return predictability approach, a methodology similar to the one of Welch and Goyal (2007), aiming to find evidence towards the predictive power of currency returns on excess stock returns. As we anticipate predictability on different time
17
horizons, we perform our forecast analysis on K=1, 6, 12, 24 and 36 month-ahead returns. The month-ahead returns are constructed as follows:
(1)
where represents k-months ahead return, is the return at time t+k, t has a monthly frequency and k represents the predictability horizon.
The first step of the methodology is the in-sample analysis. We conduct the in-sample analysis by running a single time-series regression using all available data, minimum 20 years of data, per predictability horizon per country and compare the predictive power of domestic currency returns across countries based on the regressions and coefficient significance. The regressions are run as simple regressions, but also as multiple regressions with demeaned risk-free rate as control variable. The analysis starting point is the regression formula:
(2)
where is the k-months ahead excess stock index return, is the regression constant, is the exchange rate return at time t, is the control variable of the country at time t,
the error variable and t has monthly frequency.
Thus, for the in-sample analysis we run 2 different regressions per time-horizon for each country, meaning that, for example: for the first regression we regress 1-month ahead excess stock market returns on a constant and currency returns and for the second regression we regress 1-month ahead stock market excess returns on a constant, currency returns and demeaned risk-free rate. We repeat the above for each k and for each country, ending with 200 emerging countries in-sample regressions and 50 developed countries in-sample regressions. As we conduct our forecast analysis on horizons longer than 1-month ahead, serial correlation between standard errors may lead to biased results. Hence, we adjust for such impediment by using Newey West standard errors2.
Secondly, we conduct an out-of-sample (OOS) analysis. As OOS results are improved when more data is used, we consider countries with stock and currency data available on the past 30-years to give more meaningful results when compared with countries that have less data available. There are several emerging countries that lack data availability for 30-years
2
Petersen, M. A. (2009). Estimating standard errors in finance panel data sets: Comparing approaches. The Review of
18
period: Egypt, China, Brazil, Colombia, Poland, Czech Republic and Romania; their results are still valid however they should be interpreted with more caution.
The OOS regressions have the same regression equation as the in-sample regression, but the approach is different. For in-sample regressions we use all available data, meaning that all information is already in the data. However, when conducting an OOS regression we assume that at the time of estimation we do not have all the available information, as we do not use any data available beyond the period that is estimated. We forecast the excess stock market returns on the l0 most recent years using the past available data of currency returns, typically around 20 years. As we need to run a regression for each month of the 10 years of prediction we create a loop and use an expanding window to regress excess stock market returns on exchange rate returns and controls. The regressions are then run and we create the forecast of the excess stock returns as follows: the forecast at t=241 using exchange rate returns data of the period ranged from t=1 to t=240. Then for t=242 we use data on the period range t=1 to t=241 and so on, ending with 120-k estimates of and , where k is the months ahead horizon. The initial regression equation:
(2)
where is the k-months ahead excess stock index return, is the regression constant,
is the exchange rate return at time t, is the control variable of the country at time t, the error variable and t has monthly frequency.
The forecasts are then constructed accordingly:
(3)
where is the k-months ahead forecasted return, is the exchange rate return at
time t, and are the regression coefficient estimates and t has monthly frequency.
To be able to conclude if exchange rate returns are a valid predictor of excess stock market returns we compute the using currency return data up to month t to construct the forecast for stock return t+k periods ahead and compare its prediction power to the historic mean of excess stock returns.
∑
∑
19
where is the k-months ahead excess stock index return, is the forecasted k-months ahead excess stock index return, is the historic mean of excess stock return used as a benchmark, T-k is the number of total forecasts given a time-horizon of k, in our case T=360 as we predict 10 years of data and both T and t have monthly frequency.
The intuition behind the values of is that if the historic mean of excess stock returns is a better predictor than exchange rate returns. On the other hand if then we can conclude that the exchange rate returns have certain predictive power on excess stock returns. Evidence towards emerging countries positive answers our research question but can also support our initial statement: that the difference in efficiency level between the two markets may lead to predictability.
Nevertheless, in order to prove that different market efficiency levels lead to stock returns predictability, we run a longitudinal analysis between emerging and developed countries. As emerging countries currency market is efficient and stock market is inefficient and developed countries currency market is inefficient and stock market tends to efficiency, we expect different signs between emerging and developed countries. If the prediction power of currency returns is not present in the developed countries, meaning that emerging countries
is mostly positive while of developed countries is mostly negative; we can
conclude that market efficiency plays a role in stock return predictability.
4. Preliminary Results
4.1.In-sample analysis using USD/domestic exchange rate returns as predictor
We first present the results for the in-sample analysis using USD/domestic exchange rate returns. From Table 1 it can be observed that in-sample there are several emerging countries that have predictable excess stock returns by their exchange rate returns. In the short run (K=1), the countries that show the highest signs of predictability are Philippine, Peru, Argentina, Mexico, Malaysia and Romania. Besides Peru, the effect of exchange rate returns on excess stock returns is negative, meaning that an increase in exchange rate returns (depreciation in the domestic currency) leads to lower one period ahead excess returns. For example, 1% increase in the exchange rate return of Philippines forecasts a 0.44% lower excess stock return in the following month. One reason behind this negative relation may be the “strength” of the currency. As also mentioned in the literature section, investors are
20
interested in strong currencies, thus, if a currency depreciates it may be a signal for investors that the country faces low economic performance. The effect may be amplified by the economic and financial risk of the specific countries, however the significance of the effect is beyond the purpose of this paper and it will not be described in depth in this analysis. On the other hand, the positive effect that is present in Peru may be the outcome of a cheaper currency that increases the country competitiveness on the export market. In the short run, predictability may arise solely due to the different levels of market efficiency between the currency and stock markets. As new information that is incorporated in the more efficient market prices may be incorporated in the less efficient market prices at a later date, predictability due to efficiency may be linked to how fast the less efficient market reacts to new information and to how long the information remains relevant. The intuition behind being that information relevant today has higher chances of still being relevant in 6 or 12 months rather than in 24 or 36 months. Thus, if efficiency plays a role in predictability, one would expect more predictability on shorter horizons.
Table 1
Emerging countries in-sample results using USD/domestic exchange rate
This table reports the results of the in-sample time series regressions of emerging countries excess stock returns. The dependent variable is the "k"-month’s ahead excess stock returns while the independent one is USD/domestic currency returns. Countries results are reported on separate rows. Each column indicates how many periods ahead the excess stock returns are: K=1 means that we forecast one-month ahead excess stock returns. For each country we present the regression coefficient of the independent variable, its t-statistics and the R^2. The sample contains observations on the time-period 1988-2018. Newey West standard errors are being used. Significance level is presented as follows: denotes significance at the 1% level; ** denotes significance at the 5% level; * denotes significance at the 10% level.
Country
β t-stat R^2 β t-stat R^2 β t-stat R^2 β t-stat R^2 β t-stat R^2
EGYPT -0.011 -0.12 0.01% -0.041 -0.14 0.01% -0.106 -0.21 0.02% 6.726 2.15** 2.09% 11.629 2.79*** 3.68% TAIWAN -0.237 -0.74 0.15% 0.346 0.41 0.05% 0.935 0.84 0.21% 3.424 3.01*** 2.64% 1.713 1.49 0.69% TURKEY -0.199 -1.61 0.72% 0.247 0.56 0.09% 0.569 0.83 0.20% 2.252 2.99*** 2.61% 3.530 4.12*** 5.02% CHINA -0.168 -0.69 0.15% -0.163 -0.29 0.03% -0.655 -0.8 0.21% -1.068 -1.1 0.42% 0.404 0.39 0.05% PHILIPPINE -0.443 -2.49** 1.71% -0.329 -0.66 0.12% -1.459 -1.88* 1.02% 0.809 0.8 0.19% 0.285 0.24 0.02% INDONESIA -0.001 -0.01 0.00% -0.879 -3.32*** 3.04% -1.900 -4.45*** 5.42% -0.021 -0.04 0.00% -0.054 -0.11 0.00% CHILE -0.149 -0.92 0.26% 0.034 0.08 0.00% 0.636 0.92 0.27% 1.328 1.73* 0.98% 1.609 1.69* 0.98% PERU 1.912 4.65*** 6.53% 8.930 6.87*** 13.39% 16.633 7.37*** 15.38% 17.524 7.11*** 14.96% 17.328 6*** 11.59% ARGENTINA -0.390 -3.14*** 3.03% -2.099 -5.05*** 7.60% -0.093 -0.14 0.01% 2.871 3.33*** 3.66% 3.148 3.1*** 3.32% BRAZIL -0.021 -0.2 0.01% 0.113 0.31 0.04% 0.388 0.42 0.07% 0.900 1.21 0.59% 0.593 0.84 0.30% THAILAND -0.030 -0.18 0.01% -0.431 -0.92 0.24% -1.651 -2.36** 1.59% 0.421 0.48 0.07% -0.071 -0.07 0.00% SOUTH AFRICA -0.053 -0.74 0.15% -0.030 -0.14 0.01% -0.332 -1.04 0.31% -0.240 -0.57 0.10% -0.583 -1.19 0.44% SOUTH KOREA 0.130 1.2 0.41% -0.644 -1.9* 1.02% -0.433 -0.83 0.20% 1.550 2.69*** 2.12% 1.243 2.08** 1.32% PAKISTAN -0.208 -0.68 0.14% -1.928 -2.15** 1.38% -3.040 -2.11** 1.35% -2.558 -1.38 0.61% -2.176 -0.94 0.29% MEXICO -0.446 -3.93*** 4.40% -0.974 -2.38** 1.68% -1.311 -1.9* 1.09% 1.228 1.47 0.69% 1.941 2.22** 1.61% COLOMBIA -0.049 -0.53 0.10% 0.090 0.26 0.02% 0.197 0.35 0.04% 1.536 2.22** 1.82% 1.338 1.64 1.05% POLAND -0.029 -0.21 0.02% 0.802 2.2** 1.66% 0.566 1 0.35% 1.115 1.57 0.90% 1.140 1.46 0.82% CZECH REPUBLIC -0.142 -1.27 0.58% 0.070 0.22 0.02% 0.089 0.18 0.01% -0.426 -0.65 0.17% -0.231 -0.29 0.04% MALAYSIA -0.285 -2.08** 1.20% -0.609 -1.51 0.65% -1.294 -2.03*** 1.19% 0.284 0.37 0.04% 0.076 0.09 0.00% ROMANIA -0.360 -1.92* 1.49% 0.140 0.22 0.02% 0.116 0.11 0.00% 4.935 3.1*** 4.18% 7.487 3.55*** 5.67% Forecast Horizon K=1 K=6 K=12 K=24 K=36
21
The countries for which exchange rate returns are able to predict 6-month ahead returns are Indonesia, Peru, Argentina, South Korea, Pakistan, Mexico and Poland. The outcome is similar to the 1-month ahead excess returns results, meaning that all countries but Peru have a significant negative relation while Peru has a positive one. The results for 12-months ahead are in line with the 6-months ahead as well. Significant negative effects exist in Philippines, Indonesia, Thailand, Pakistan, Mexico and Malaysia while the only significant positive effect is current in Peru. In the long-run a shift occurs. By looking at the 24-months ahead and 36-months ahead returns one can observe that all significant relations are positive. Egypt, Taiwan, Turkey, Chile, Peru, Argentina, South Korea, Colombia and Romania are the countries that stand out when 24-months ahead returns are predicted and Egypt, Turkey, Chile, Peru, Argentina, South Korea, Mexico and Romania are the countries that have a significant relation between exchange rate returns and 36-months ahead excess stock returns. The long run positive relation may arise due to the increased competitiveness of the countries on the foreign markets as a result of the currency depreciation: e.g. domestic products are cheaper to buy on the export market. Overall, we find in-sample predictability in 16 out of the 20 chosen emerging countries excess stock returns.
We find predictability in-sample, however because emerging countries are subjects of economic instability, historically, their risk-free rate tends to suffer dramatic shifts. For example, in 1990 the deposit interest rate in Brazil reached an all-time high of 9394.29%3. To check if the results are robust to these risk-free rate shifts we perform similar regressions to the ones presented in Table 1 and add a demeaned risk-free rate variable as control. By adding the demeaned risk-free rate variable, we are able to control the risk-free rate swings and check if exchange rate returns are still being able to predict excess stock returns. The outcome is shown in Table 2.
If we compare the results in Table 1 to the ones in Table 2 it can be noticed that the countries that have significant results are rather similar. The direction of the effect is alike. Nevertheless, few countries excess stock returns lose their predictability while other countries excess stock returns become significantly predictable. In the short-run (K=1) predictability results are robust in Philippine, Argentina, Mexico and Romania. On the other hand, Peru and Malaysia exchange rate returns lose their predictive power while South Korea excess stock returns become predictable. The 6-months ahead excess stock returns remain predictable in
3
22
Indonesia, Peru, Argentina and Pakistan whereas Egypt returns become predictable when controlled with the demeaned risk-free rate. On the same time-horizon, Poland and Mexico excess stock returns lose their predictability. The 12-months ahead excess stock returns remain predictable only for Indonesia, Peru, Thailand and Pakistan. On longer time-horizons excess stock returns predictability by exchange rate returns remains unchanged between no-control and no-controlled with the demeaned risk free rate regressions: the exact same countries have significantly predictable 24 and 36-months ahead excess stock returns by exchange rate returns. Overall, compared with the non-controlled regressions, the number of countries with significantly predictable excess stock returns has decreased by 2, Poland and Malaysia stock returns losing their predictability when controlled for demeaned risk-free rate. One noticeable difference between the results in Table 1 and Table 2 are the values. measures the percentage of the dependent variable explained by the independent variables. Table 2 presents slightly higher than Table 1 as the regression equation used in constructing Table 2 uses two independent variables while Table 1 regression equation contains only one.
Table 2
Emerging countries in-sample results using USD/domestic exchange rate and demeaned risk-free rate
This table reports the results of the in-sample time series regressions of emerging countries excess stock returns. The dependent variable is the "k"-month’s ahead excess stock returns while the independent ones are USD/domestic currency returns and the domestic demeaned risk-free rate. As the demeaned risk-free rate is used for control purposes its coefficients are not shown. Countries results are reported on separate rows. Each column indicates how many periods ahead the excess stock returns are: K=1 means that we forecast one-month ahead excess stock returns. For each country we present the regression coefficient of the independent variable, its t-statistics and the R^2. The sample contains observations on the time-period 1988-2018. Newey West standard errors are being used. Significance level is presented as follows: denotes significance at the 1% level; ** denotes significance at the 5% level; * denotes significance at the 10% level.
Country
β t-stat R^2 β t-stat R^2 β t-stat R^2 β t-stat R^2 β t-stat R^2
EGYPT 0.035 0.38 2.29% 0.222 0.76 5.63% 0.661 1.32 8.99% 5.836 1.93* 9.82% 10.897 2.65** 6.79% TAIWAN -0.324 -1.06 2.50% -0.066 -0.08 16.26% 0.318 0.32 21.57% 2.706 2.65*** 10.75% 0.668 0.66 19.95% TURKEY 0.076 0.65 19.11% 0.948 2.18** 9.75% 0.935 1.34 1.86% 2.243 2.97*** 2.67% 3.657 4.05*** 5.07% CHINA -0.125 -0.52 2.39% 0.086 0.16 13.71% -0.247 -0.33 17.30% -0.653 -0.72 13.78% 0.780 0.78 8.91% PHILIPPINE -0.392 -2.16** 2.51% 0.081 0.16 6.05% -0.718 -0.94 8.70% 1.127 1.12 1.31% 0.267 0.22 0.02% INDONESIA 0.053 0.9 26.31% -0.688 -2.99*** 24.41% -1.710 -4.21*** 12.89% -0.093 -0.18 1.00% -0.109 -0.22 0.25% CHILE -0.166 -1.15 20.21% 0.008 0.02 8.13% 0.573 0.92 19.08% 1.291 1.73* 6.67% 1.546 1.7* 10.06% PERU 0.226 0.81 60.63% 3.800 4.12*** 60.37% 7.179 4.96*** 68.28% 8.960 4.62*** 52.12% 7.689 3.27*** 46.89% ARGENTINA -0.235 -2.19** 29.81% -1.785 -4.49*** 17.20% 0.319 0.49 6.93% 2.985 3.43*** 3.98% 3.170 3.09*** 3.33% BRAZIL -0.021 -0.2 0.06% 0.111 0.31 1.21% 0.363 0.63 4.72% 0.896 1.24 6.61% 0.588 0.86 5.83% THAILAND 0.046 0.28 6.22% -0.105 -0.24 12.39% -1.320 -1.95* 7.81% 0.590 0.66 1.26% 0.380 0.35 4.89% SOUTH AFRICA -0.005 -0.07 14.70% 0.160 0.85 24.48% -0.092 -0.32 20.61% 0.069 0.18 17.36% -0.347 -0.72 7.79% SOUTH KOREA 0.362 3.6*** 20.86% 0.336 1.15 36.68% 0.564 1.09 15.87% 1.811 2.99*** 3.58% 1.652 2.73*** 5.33% PAKISTAN -0.054 -0.18 8.43% -1.791 -2.09** 12.15% -2.492 -1.77* 11.11% -1.576 -0.83 5.79% -1.457 -0.62 5.97% MEXICO -0.275 -2.49** 14.83% -0.270 -0.69 15.63% -0.268 -0.4 11.86% 1.273 1.48 0.70% 2.223 2.46** 2.11% COLOMBIA -0.105 -1.33 26.74% -0.125 -0.43 28.72% -0.019 -0.04 11.01% 1.597 2.31** 2.42% 1.430 1.75* 1.92% POLAND -0.080 -0.59 4.74% 0.483 1.56 29.54% 0.021 0.05 38.10% 0.605 0.95 21.75% 0.654 0.91 17.22% CZECH REPUBLIC -0.155 -1.4 3.69% 0.007 0.02 12.19% -0.007 -0.02 14.05% -0.477 -0.73 2.38% -0.298 -0.38 2.87% MALAYSIA -0.177 -1.35 12.44% -0.221 -0.6 20.06% -0.795 -1.34 16.19% 0.455 0.58 1.61% 0.334 0.39 2.12% ROMANIA -0.354 -1.99* 11.53% 0.171 0.28 10.60% 0.181 0.17 1.95% 4.968 3.13*** 4.92% 7.604 3.66*** 9.04% K=36 K=1 K=6 K=12 K=24 Forecast Horizon
23
As this paper aims to find if exchange rate returns are being able to predict excess stock market returns but also prove that predictability is stronger in emerging countries when compared with developed countries due to the difference in market efficiency, the same in-sample analysis is done on a in-sample of developed countries. Table 3 is the replica of Table 1 using developed countries instead of emerging ones. The results suggest predictability in the short to medium-run but no predictability in the long-run. The countries that have predictable excess stock returns at 1-month horizon are United Kingdom and Sweden. The 6-months ahead excess stock returns are predictable with exchange rate returns only in Sweden. The longest horizon on which excess stock returns are predictable in-sample for developed countries is 12-month ahead: Australia and Sweden have significantly predictable excess stock market returns. One important difference between emerging and developed stock markets in-sample results are the signs of the significant β coefficients: an increase in emerging countries exchange rate returns (depreciation of domestic currency) lead to a decrease in “k” months ahead excess stock market returns on short to medium-run while for developed countries an increase in exchange rate returns (depreciation of domestic currency) leads to an increase in “k” months ahead excess stock market returns on short to medium-run. There may be several reasons behind this difference, stock market development, investors experience, country’ economic and financial risk: as developed countries are perceived less risky investors may consider the depreciation of the domestic currency an “aggressive” competitive behaviour of the country rather than economic instability.
Table 3
Developed countries in-sample results using USD/domestic exchange rate
This table reports the results of the in-sample time series regressions of developed countries excess stock returns. The dependent variable is the "k"-month’s ahead excess stock returns while the independent one is USD/domestic currency returns. Countries results are reported on separate rows. Each column indicates how many periods ahead the excess stock returns are: K=1 means that we forecast one-month ahead excess stock returns. For each country we present the regression coefficient of the independent variable, its t-statistics and the R^2. The sample contains observations on the time-period 1988-2018. Newey West standard errors are being used. Significance level is presented as follows: denotes significance at the 1% level; ** denotes significance at the 5% level; * denotes significance at the 10% level.
Country
β t-stat R^2 β t-stat R^2 β t-stat R^2 β t-stat R^2 β t-stat R^2
JAPAN 0.038 0.28 0.02% -0.165 -0.49 0.07% -0.168 -0.33 0.03% -0.305 -0.45 0.06% -0.649 -0.79 0.19% CANADA -0.005 -0.04 0.00% 0.057 0.19 0.01% 0.097 0.23 0.02% 0.355 0.71 0.15% 0.264 0.51 0.08% AUSTRALIA 0.017 0.25 0.02% 0.173 0.9 0.24% 0.816 2.77*** 2.26% 0.604 1.63 0.83% 0.032 0.07 0.00% UK 0.227 2.56** 1.81% 0.160 0.75 0.16% 0.550 1.69* 0.82% 0.540 1.22 0.44% 0.000 0 0.00% SWEDEN 0.425 3.12*** 2.82% 1.144 2.9*** 2.49% 1.365 2.19** 1.46% 0.436 0.52 0.09% -0.359 -0.36 0.04% Forecast Horizon K=1 K=6 K=12 K=24 K=36
