Bachelor Thesis in Economics May 2011 Stock returns explained - using a volume filter, interest rates, and the oil price Supervisors: Author: Hossein Asgharian Pierre R.M. Carlsson Title: Stock returns explained - using a volume filter, interest rates, and the oil price. Seminar date: 2011-05-31 Course: Bachelor thesis in Economics, 15 ECTS Author: Pierre R.M. Carlsson Advisor: Hossein Asgharian, Department of Economics Key Words: Econometrics, Information asymmetry, Interest rate, Oil price, and Volume. Purpose: Account for trade volume activity and investigate the explanatory power of a few generally accepted variables ability to explain index and stock returns. Empirical The Swedish stock index OMXS30 and constituents has foundation: empirically been studied to obtain the data needed. Theoretical The theory is derived from research on macroeconomic variables perspective and the price-volume relation. By combining them value is added. Methodology: A quantitative approach using regression analysis have been used. Conclusions: Filtering for volume provides additional insights of when an explanatory variable is useful. It further provides insights that the sign and size of these impacts could vary, significantly, depending on the trade volume activity. The most reliable and consistent variables was the oil price, showing a positive relation, followed by the term spread, also positive. The results further demonstrate significant differences between high and low turnover stocks. Acknowledgements First of all I would like to recognize my parents who are always there for me. Thank you for always letting me walk my own way. Since this is my last major coursework at Lund University I would also like to take the opportunity to thank everyone at Lund University and all friends I have made here. Thank you for having made my time in Lund the very best. Third, I would like to thank my supervisor, Hossein Asgharian, for the brilliant knowledge and wisdom you have taught me over the years. Last but not least I would like to thank the city of Lund for a great city to live, study and work in. Abstract Using a volume filter on daily index and stock price data the daily return has been researched. The explanatory variables used in the study are the 1 M T-Bill, the term spread - 10 Y Treasury bond versus a 3 M T-Bill -, and the oil price. The results revealed that accounting for trade volume is an important part in explaining the return of a stock or index. The volume activity provides additional insights of when a relation between the explanatory variables and the stock return are valid. It also reveals that the relation varies significantly across different volume activity. The most reliable and consistent variables was the oil price and the term spread, both demonstrating a positive relation. The results also revealed that there are differences between high and low turnover stocks. Table of Content 1 2 3 4 INTRODUCTION ................................................................................................ 1 1.1 Background ............................................................................................................................. 1 1.2 Problem discussion ................................................................................................................. 3 1.3 Purpose .................................................................................................................................... 5 1.4 Delimitations ........................................................................................................................... 5 1.5 Thesis outline .......................................................................................................................... 5 THEORETICAL FRAMEWORK ..................................................................... 6 2.1 Volume .................................................................................................................................... 6 2.2 Interest rate ............................................................................................................................. 9 2.3 Oil price ................................................................................................................................. 11 METHODOLOGICAL FRAMEWORK ......................................................... 13 3.1 Research design .................................................................................................................... 13 3.2 Data selection ........................................................................................................................ 15 3.3 Construction of explanatory variables ............................................................................... 15 3.4 Regression model .................................................................................................................. 16 3.5 Hypothesis discussion ........................................................................................................... 17 3.6 Methodological difficulties ................................................................................................... 19 EMPIRICAL RESULTS ................................................................................... 20 4.1 The OMXS30 index .............................................................................................................. 20 4.2 The OMXS30 index constituents ......................................................................................... 23 4.3 Concluding remarks and main results ................................................................................ 27 5 ANALYSIS AND DISCUSSION ....................................................................... 28 6 CONCLUSION ................................................................................................... 33 6.1 Criticism of research ............................................................................................................ 33 6.2 7 Further studies...................................................................................................................... 34 REFERENCES ................................................................................................... 35 APPENDIX ................................................................................................................. 45 Table 1 –Volume characteristics for different volume groups of OMXS30 ............... 20 Table 2 – Return characteristics for different volume groups for OMXS30 ............... 21 Table 3 – Correlation matrix ........................................................................................ 21 Table 4 – Regression results from the OMXS30 index ............................................... 22 Table 5 –Summary of the volume within different volume groups for all firms ......... 23 Table 6 –Summary of the return within different volume groups for all firms ........... 24 Table 7 – Summary of the regression results for ONLY significant variables among all firms ........................................................................................................................ 25 Table 8 – Hypothesis table, OMXS30 index ............................................................... 28 Table 9 – Hypothesis table, OMXS30 index constituents, relation and significance .. 29 Table 10 –Firms in the study with corresponding sector, industry group and subindustry (GICS) ............................................................................................................ 45 Table 11 – Descriptive statistics of firms sorted by turnover (1 of 2) ......................... 46 Table 12 – Descriptive statistics of firms sorted by turnover (2 of 2) ......................... 47 Table 13 – Descriptive statistics of firms sorted by sector (1 of 3) ............................. 48 Table 14 – Descriptive statistics of firms sorted by sector (2 of 3) ............................. 49 Table 15 – Descriptive statistics of firms sorted by sector (3 of 3) ............................. 50 Table 16 – Results from regression model for firms sorted by sector, 1 of 2 (read together with table 17 and 21) ..................................................................................... 51 Table 17 – Results from regression model for firms sorted by sector, 2 of 2 (read together with table 16 and 21) ..................................................................................... 53 Table 18 – Results from regression model for firms for high turnover, 1 of 2 (read together with table 19 and 20) ..................................................................................... 54 Table 19 – Results from regression model for firms for low turnover, 2 of 2 (read together with table 18 and 20) ..................................................................................... 55 Table 20 –Results associated to the regression models for firms sorted by turnover (see 18 and 19 ) ............................................................................................................ 56 Table 21 –Results associated to the regression models for firms sorted by sector (see 16 and 17) .................................................................................................................... 56 Table 22 - The average value for each explanatory variable in the three volume groups ........................................................................................................................... 57 Table 23 – OMXS price graph with the return, and high and low volume days ......... 58 Table 24 – A graph of the explanatory variables ......................................................... 59 1 Introduction In this introductory chapter choice and motives behind the research topic are presented and this leads up to the purpose of the thesis. The chapter is ended by delimitations and a disposition of the thesis. 1.1 Background Several studies have revealed that past trading volume contains information not accounted for in the past stock return. (Gervais et al., 2001) Moreover, intuition and research have concluded that the market consists of different investors; institutional and individual, which further interprets information differently, and trade accounts of different size. Therefore, in trying to explain, and predict, stock returns the best results would emerge if one could capture and model each investor group separately. More accurate information could be discovered in doing so. In capturing the behaviour of each investor group one could potentially acquire better models with higher predictive ability, which exogenous factors that are important in explaining the return, and an overall improved foundation for intelligent decision making. This paper will be a first effort in trying to separate the high volume, suggestively consisting of a high share of institutional volume, from the low volume, containing a suggestively low share of institutional investors. The rationale behind separating the different trade volume activity is that the information content in high (low) volume periods is concluded to be positive (negative) and very robust. It has further been suggested that this remains in effect independent of how trading volume is measured, if it is adjusted for firm announcements, return effects, and it has been viable for several decades. (Gervais et al., 2001; Lee and Swanminathan (1998) among others) The institutional investor is expected to be better informed than the individual investor, as well as demonstrate more investor intelligence. The institutional would by experience automatically filter away a lot of noise in the economic, the firm and the graph specific information/announcements. It is therefore important to distinguish between these two major investor groups. 1 In summary, different investors do not necessarily “see” the same firm fundamentals or future which contributes to different actions and behaviour. This in turn, could potentially be extracted from the past trade volume activity. 2 1.2 Problem discussion In the financial market different investors has, naturally, different access to information. This information asymmetry is one contributing factor to the framework suggested by the efficient market hypothesis (EMH). According to the EMH there are three different efficiency levels: weak efficiency, the price already reflects all past publicly available information, semi-strong efficiency, the price reflect all publicly available information and instantly change to reflect new, and strong efficiency, price instantly reflect private information. In other words, no information advantage and information advantage. When one has (no) information advantage he/she is expected to conduct in (in)significant trade activity. This give rise to, on average, high volume days with suggestively a high share of institutional investors, average volume days consisting of a more symmetric mix of investors, and third, low volume days with a high share of individual investors. However, using just daily historical volume records to determine the exact ratio between institutional and individual investors is not possible. On the other hand, it is impossible for individual investors to consistently generate days with high trading volume. That said, it is probable that an individual investors have an information advantage, and as this spreads to the investment community the flow of information could act as a trigger for the high volume to enter the market. Altogether, using volume as a filter one should approximately be able to separate the days with high institutional volume activity from days with low institutional volume activity (a high share of individual volume activity). In addition to information asymmetry and the information contained in the past volume activity several variables and ratios have been used and developed to gain insights about the future stock return. The two major groups categorizing most of them are financialand macroeconomic variables. Starting with the financial variables, such as market capitalization, cash flow yield, dividend yield, foreign exchange, earning-to-price ratio, price-to-book ratio, and turnover etc. several authors have analyzed them trying to find the most useful and significant variables. (Fama and French, 1992; Chan et al., 1991; Banz,, 1981; Basu, 1983; Litzenberger and Ramaswamy, 1982; Adler and Dumas, 1983; Roll, 1992; Dumas and Solnik, 1995; Clasessens et al., 1995; Keim and Stambaugh, 1986; Pontiff and Schall, 1998; Lewellen, 2004) However, there is limited consensus on 3 which variables are consistent predictors of equity return and when one should use them. In addition, these ratios cannot always be calculated for all firms, e.g all firms do not pay dividend etc., which is why they will not be considered in this study. Shifting to the macroeconomic variables, such as aggregated output, default spread, imports and exports, inflation rate, industrial production, interest rates, money stock, term spread, and unemployment rate etc. the general consensus is that stock returns are predictable using macroeconomic variables (Ang and Bekaert, 2001; Balvers et al., 1990; Bodie, 1976; Campbell, 1987, 1990; Chen et al., 1986, Chan et al., 1998; Chen, 1991; Chen, 2009; Conover et al., 1999; Cutler et al., 1989; Fama, 1981; Fama and Schwert, 1977; Fama and French, 1989; Ferson and Harvey, 1993; Flannery and Protopapadakis, 2002; Geske and Roll, 1983; Hodrick, 1989; Jaffe and Mandelker, 1976; Lamont, 2001; Nelson, 1976; Pearce and Roley, 1983, 1985; Pesaran and Timmermann, 1995; Rapach et al., 2005) That said, consistent with the financial variables there is limited consensus on which variables are robust predictors of stock returns. As regards of the macroeconomic variables the interest rate appears to be the most useful one followed by, to a lesser extent, inflation. (Rapach et al., 2005 and Chen, 2009) An alternative measure/proxy to inflation, and economic activity, is oil. Oil is the fundamental driver of modern economic activity and various studies have shown that the oil price have a significant effect on the macro economy; GDP growth, inflation, and the stock market. (Apergis and Miller, 2008; Chen, 2010; Driesprong et al, 2008; El-Sharif et al., 2005; Faff and Brailsford, 1999; Gisser and Goodwin, 1986; Hamilton, 1983; Huang et al, 1996; Jones and Kaul, 1996; Kilian and Park, 2009; Keane and Prasad, 1996; Lescaroux and Mignon, 2008, Lardic and Mignon, 2006, 2008; Mork, 1989; Mory, 1993; Mork et al., 1994; Mussa, 2000; Nandha and Faff, 2008; Park and Ratti, 2008; Sadorsky, 1999;) Hence, its high dependence in our society and driver of the economy seems to make it an acceptable proxy to account for inflation and contribute to explaining stock returns. In summary, research have documented that different level of trade activity contain different information content. Together with the fact that the market is made up of 4 differently informed investors one could suspect that classical exogenous variables is more viable under certain volume characteristics. 1.2.1 Research questions i. Do interest rates and the oil price explain the return of index and stocks? ii. Investigate if the above mentioned exogenous variables have different explanatory power under different volume activity. iii. Study if there are differences between high and low turnover stocks and sectors among the firms listed in OMXS30, the major Swedish stock index. 1.3 Purpose The purpose is to account for the daily trade volume activity and investigate the explanatory power of the oil price, the term spread, and a short interest rate and their ability to explain index and stock returns. 1.4 Delimitations The research is conducted on the Swedish OMXS30 index and its constituents. The study relies upon data from 1991-01-01 to 2011-04-29. 1.5 Thesis outline The rest of the thesis is organized as follows. Chapter 2 gives an overview of the, for the thesis, relevant theoretical framework with focus on the selected variables. Chapter 3 discusses the methodology and data collection. In Chapter 4 the empirical findings from the study is described in text and tables. Chapter 5 contains an analysis and discussion of the empirical findings. Chapter 6 concludes the thesis accompanied by suggestions for future research. 5 2 Theoretical Framework In this chapter I present the theoretical context for the study. It narrows down into a summary accompanied by a few hypotheses at the end of each section. 2.1 Volume 2.1.1 Volume and return It is apparent that institutions trade in larger sizes than individuals, and research has further suggested that institutions are better informed and/or more sophisticated than individuals. (Cohen et al., 2002; Daniel et al., 1997; Holden and Subrahmanyam, 1992; Nofsinger and Sias, 1999) Consequently, investigating the trade volume could reveal information about future stock return. An extensive body of research have conducted an investigation on aggregated trade volume and stock return, and found a relation. (Gallant et al., 1992; Karpoff, 1987; Llorente, et al., 2002; Schwert, 1989) Shu, 2010 distinguish the trading volume from institutional and individual investors and show that the allocation of volume has significant impact on stock returns. The results reveal that stocks with lower fractions of institutional trading volumes underperform stocks with higher institutional volume. Chordia and Swaminathan (2000) find that daily and weekly returns on high volume portfolios lead returns on low volume portfolios. They argue that these patterns arise because returns on low volume portfolios respond more slowly to information in the market than high volume portfolios. The spread of information in the market was examined by Michael and Starks (1988). They investigate the relationship between stock prices and volume using the Granger causality1 technique and found that information is processed by investors sequentially rather than simultaneously or all at once. This finding was consistent with Copeland (1976) sequential arrival of information model in which information is disseminated to only one trader at a time and that implies a positive correlation between volume and change in price. Hence, past trading volume should hold information that is not, always, fully incorporated in the price. This conclusion was 1 A statistical tests used to determine whether one time-series is good at forecasting another. 6 reached by Lee and Swaminathan (1998) which argue for an “investor expectation hypothesis” in which past trading volume is a proxy for the level of investor interest in a stock. Low-volume stocks (low investor interest) have a greater upside potential on average, while high-volume stocks (high investor interest) face greater risk. They also provide an explanation arguing that investors holding illiquid stock demand a return premium. An extensive amount of research has further studied volume effects in relation to news announcements and price movements. According to Easley and O´hara (1992) and Bessembinder and Segin (1993) an unusual high or low volume are potential signs of the arrival of new information. Stickel and Verrecchia (1994) reason that as volume increases, the probability that the price change is information driven increases. This seems to be the evidence as large price changes on days with weak volume support tend to reverse the next day. While large price increases with strong volume support tends to be followed by another price increase the next day. (Abbondante, 2010) This was also found by Gervais et al. (2001) which show that stocks experiencing unusually high trading volume over a day or a week tend to appreciate over the course of the following month. Their rationale for this investor behavior appears to be that the high-volume return premium is consistent with the idea that shocks in the trading activity affects its visibility and in turn the subsequent demand and price for that stock. Their results also indicate that the flow of information is related to volume and not the occurrence of news events, contrary to Easley and O´hara (1992) among others. 2.1.2 Volume and volatility According to Poon and Granger (2003) several characteristics of financial markets volatility have been documented. Some among the many are volatility clustering, in which volatility vary over time where high (low) absolute return are followed by high (low) absolute returns (Mandelbrot, 1963), asymmetric reactions to shocks, in which the volatility of returns increases more following negative shocks than positive shocks of equal size (Black, 1976), fat tailed distributions of risky asset returns and long memory of volatility. (Mandelbrot and van Ness, 1968) 7 Models by Clark (1973), Epps and Epps (1976), and Tauchen and Pitts (1983) concluded that there is an existence of a positive contemporary relationship between volatility and trading volume. Lamoureax and Lastrapes (1990) argues that the daily trading volume is a measure of the amount of information flowing into the market every day. They find that trading volume seems to be a good proxy for the arrival of information into the market and for explaining the persistence of the volatility of the return of individual shares. In other words, ARCH2 effect tend to disappear when contemporaneous trading volume is added to the conditional variance function of a GARCH(1,1)3 specification. Fleming et al. (2005) study the degree to which the dynamics of trading volume can explain ARCH in stock returns. In contrast to previous authors they find that trading volume, inserted into the conditional variance function, do not reduce the importance of lagged squared returns in capturing volatility dynamics. They use a EGARCH4 (2,2) model that allows for both short- and long-term volatility components and find little support for the proposition that volume explains ARCH effects. However, the model does imply that volume is strongly associated with return volatility. 2.1.3 Concluding remarks of volume It has been concluded that institutional investors trade in larger size and are more sophisticated as they account for more information. It is further suggested that high volume activity lead as these investors behavior act as a signal to other investor groups. Conditional on volume the return is further expected to exhibit different characteristics, such as higher volatility for high volume than low volume stocks. Presented that high volume incorporate more information than low volume the explanatory variables should provide better explanatory power on high volume days than on low volume days. It is further possible that volume clustering will be found, which some argues correlates with the well documented return volatility clustering. (Clark, 1973; Chordia and Swaminathan, 2000; Gervais et al., 2001; Karpoff, 1987; Lee and Swaminathan, 1998) 2 Engle, (1982) (ARCH, Autoregressive Conditional Heteroskedasticity, a method used to account for timevariation in the past error term in time-series analysis. 3 Bolleslev, (1986), (GARCH, Generalized ARCH, a method used to, in addition to ARCH, also account for time-variation in the past variance 4 Nelson, (1991), EGARCH, the exponential-GARCH allow for asymmetries between the volatility and the return, and it is the log of variance is used, which consequently will always be positive. 8 2.2 Interest rate Macro variables such as the term structure of interest rate are commonly associated with expectations of future economic events that may affect the stock market. (Chen, 2009) Avramov and Chordia (2006) conduct further research on the topic and find that returns are predictable out-of-sample by the term spread (the difference between treasury bonds with >10 years to maturity and the T-bill that matures in three months) and the one month T-bill yield, among other variables which are not as robust. Several authors have used a 3 month T-Bill, one of them are Perez-Quiros and Timmermann (2000) which uses it as a proxy for investors’ expectations of future economic activity. They also argue, consistent with others5 that it is a proxy for firms’ interest costs and find it to be negatively correlated with future returns. In a larger study, accounting for a wide range of macroeconomic variables across twelve industrialized countries, Rapach et al., (2005) find that the relative6 3-month treasury bill rate, the relative4 long term bond yield, the relative4 money market rate, demonstrate the most consistent and reliable in-sample and out-of-sample predictors of stock returns. As regards of the term spread they find limiting evidence for this to predict stock returns. However, the usefulness of term spread has been investigated by Estrella and Mishkin (1998) which found it to be a good predictor of recessions (both in-sample and out-of-sample). The discovery that some macro variables are good at predicting bear markets has been given further attention. Chen (2009) concluded that macroeconomic variables, and the term spread in particular, are better predicting bear markets rather than stock returns. Chang (2009) account for the fact that bear and bull markets should be treated separately. This is taken into account as the interest rate, dividend yield and default premium is analyzed on US stock return movements using a regime switching model. The results show that stock returns and volatility depend on macro factors and the degree of influence do change with the stock market conditions. It is also concluded that the contribution of default premium and interest rate to stock returns is significantly greater in a volatile regime compared to that in a stable regime. 5 6 Fama and Schwert, 1977; Campbell, 1987; Glosten et al., 1993; and Whitelaw, 1994 Defined as the difference between itself and its 12 month moving average 9 Besides Chen (2009) finding that the term spread predicts bear markets it was also found that the inflation rate do so. Regarding the relationship between the interest rate and inflation, Feldstein and Eckstein (1970) concluded that a one percent increase in the interest rate was approximately the result of a one percent increase in the anticipated inflation. A theory trying to explain the relationship between inflation and interest rate is the Fisher theory. Put into a stock market context the Fisher theory suggest that the relation between stock returns and inflation should be positive. (Patro et al. 2002) In contrast Fama (1981) argues and find support for, using US data, that an increase in inflation is expected to be followed by a decline in real economic activity and corporate profits. Hence, stocks will react negatively to a rise in inflation. According to the study by Rapach et al. (2005) the inflation rate demonstrate a significant in-sample and out-ofsample stock return predictive ability. However, the low frequent nature of inflation rate data makes it a less attractive variable for this study. 2.2.1 Concluding remarks of interest rates Altogether, the variables which have demonstrated the most reliable and consistent explanatory power of the interest rate variables are a short rate, the term spread and the inflation. Therefore, according to previous research the one month T-bill should be a good measure to reflect the expectations of future economic activity, influences on volatility and the cost of capital. It is further expected to be negatively related to stock returns. (Perez-Quiros and Timmermann, 2000) A second interest rate variable which has been considered useful is the term-spread. It will be used as measure to capture the state of the economy. The term spread is expected to demonstrate a positive relation to stock returns. (Chen, 2009) To account for the low frequent data of inflation the oil price will be considered a possible proxy. 10 2.3 Oil price According to research investigating the consumer price index several studies have concluded that an oil price increase represents an inflationary shock. (Fuhrer, 1995; Gordon, 1997; Hooker, 2002) Barsky and Kilian (2004) show that oil price increases generate high inflation while LeBlanc and Chinn (2004) conclude that it has only a moderate impact on inflation. The advantage of accounting for oil is that it is the one fundamental driver of modern economic activity in our world today. It is therefore concluded that the oil price should be an acceptable proxy for inflation. Different studies have investigated the relation between oil price movements on gross domestic product and on prices. The general conclusion has been that rising oil prices leads to a reduction of potential output (Brown and Yücel, 1999, 2002; Hamilton, 1983, 2005; Gisser and Goodwin, 1986; Mussa, 2000). According to available research it has been demonstrated that the impact of oil price changes on the macro economy is asymmetric. (Brown and Yücel, 2002; Ferderer, 1996; Lardic and Mignon, 2006, 2008; Mork, 1989; Mork et al., 1994; Mory, 1993) Lescaroux and Mignon (2008) extend the analysis and investigate various links between oil prices and several macroeconomic and financial variables. Their short-term analysis indicates that when Granger-causality exists, it generally runs from oil prices to other variables. Their long-term analysis reveals that GDP and oil prices evolve together (for 12 countries). According to Lescaroux and Mignon a rise in energy prices causes a drop in productivity, which is passed on to (i) real wages and employment; (ii) selling prices and core inflation; and (iii) profits and investments, as well as stock market capitalization. (Brown and Yücel, 2002 and Lardic and Mignon, 2006) Caruth et al. (1998) and Davis and Haltiwanger (2001) investigated the impact of oil price movements on the labour market and the natural rate of unemployment. Their results, consistent with Keane and Prasad (1996), conclude that the impact of oil price movements can differ with the considered horizon; in the short run prices tend to reduce employment, but in the long run it tend to increase it. 11 Several authors have studied the link between oil prices and the stock market. According to Nandha and Faff, 2008 there is a common market perception that stock markets react to oil price shocks. And Sardosky (1999) has concluded that the oil price influence share prices. Driesprong et al. (2008) discover that changes in oil prices predict stock market returns worldwide, and that an oil price increase drastically lowers future stock returns. Nandha and Faff (2008) show that oil price rises have a negative impact on stock returns for all 35 global equity industry sectors but mining, oil and gas they investigate. Jones and Kaul (1996) argue that the oil prices impact the US stock market through its influence on expected dividends and cash flows. In the study by Lescaroux and Mignon (2008) their analysis reveals that there exists a strong negative Granger-causality running from oil prices to the stock market and share price. And for almost every country in their study oil prices are found to lead countercyclically share prices. Altogether the oil price appears to be a good proxy for inflation as well as be able to capture the economic activity. However, as Lescaroux and Mignon points out, our dependence on oil today is not as high as it was some decades ago, suggesting a lower impact of the oil price on economic activity than previously throughout our history. 2.3.1 Concluding remarks of oil prices Research on the oil price has demonstrated an inverse relation to dividend, cash flows and corporate profits. Given societies high dependence on oil it will primarily be used as a proxy for inflation, but also reflect the economic activity. Hence, a negative impact on stock returns is one possible relation. On the contrary, if firms successfully adjust their prices to account for higher oil prices it could demonstrate a positive relation on stock returns. (Patro et al. 2002, Driesprong et al., 2008 and Nandha and Faff, 2008) Also, if oil prices are increasing the global economy is booming, contributing to increased corporate profit across sectors which suggest a positive relation. 12 3 Methodological framework In this third chapter I give a description of the methodology used in order to perform my proposed research. I describe how data for the study was collected and constructed. Towards the end of the chapter a discussion and formulation of hypotheses are presented. 3.1 Research design Research design is a framework for gathering and analysis of data. The choice of research design reflects the stands the researcher have taken regarding what priority be given to the number of dimensions and aspects in the research process (Bryman and Bell, 2003) 3.1.1 Research philosophy and research approach. The research philosophy is associated with the view that was taken on the research process. The philosophy captures the way the researcher view the world and subsequently affects the research design, the data collection and the analysis of the study (Saunders, et al., 2003). In this study I will focus on objective and quantifiable observations that can be analysed and result in consistent and regular generalizations. Therefore, certain limitations were made already in the introductory chapter. Research approach can be described as the theoretical design of the research. Of the two main approaches, the one that is best suited for my study is the deductive one. The deductive approach has a structured design in which existing theories are examined through hypothesis testing (Saunders et al., 2003), which fit best with how I want to carry out my study. The study is an extension of previous econometric research on stock return and its relation to the information content provided by the historical price volume records and macroeconomic variables. This research is a complement to Lee and Swaminathan (1998) work on the relation between volume and return, Rapach et al. (2005) study using macroeconomic variables to explain/predict stock returns, and Nandha and Faff (2008) research investigating the relation between oil prices and stock returns, among others. The results should be of interest for those with a passion for indices and stocks and add to their expertise. 13 3.1.2 Reliability, validity, and generalisation It is of importance that the research results are trustworthy and reliable. Reliability is often of concern in quantitative research since the researcher is interested in whether the measurement technique is stable or not. A high degree of reliability is assured if it were to generate the same results if performed again. (Bryman and Bell, 2003) For a study to have validity it should measure what it sets out to measure i.e. does the data really measure what the authors intended and can the conclusions drawn from the study actually be made based on it? (Bryman and Bell, 2003; and Saunders et al, 2003) The research depends on data from highly accredited sources. Moreover, the software packages used, MS Excel 2007 and Eviews 7.0, are commonly used by the academic community as well as industry practitioners. The econometric methodology follows standard statistical procedure used by practitioners researching financial data. As mentioned previously it is of course impossible to surely conclude which investors, institutional or individual, contribute to the daily volume on any particular day. On the other hand it is intuitive that only the sophisticated investors, institutional, would have the ability to contribute to the above normal volume. When they are less active it would result in a below normal volume. Relying on close to 20 years of daily data for most stocks potential bias and a miss-categorization of investors within the different volume groups are reduced significantly. Since all the explanatory variables that have been used are general for an economy (also publically available) it should be possible to generalize the results to other stocks. 14 3.2 Data selection The research accounted for in this research is found up to 85% using the Electronic Library Information Navigator (ELIN) and the remaining 15 % are contributed to the Social Science Research Network (SSRN). The data this thesis is built upon is secondary data from Nasdaq OMX Nordic’s website (stock data), The Swedish Centralbank, Riksbanken’s website (interest rates data) and Datastream Advance (OMXS30, oil price and USDSEK data). The data used are daily observations. Adjusted data for OMXS30 and its constituents are used. 3.3 Construction of explanatory variables R = the first difference in the log of the daily price. TB = the first difference in the log of the 1 month T-bill, SSVX 1M7. TS = term spread, the difference between treasury bonds with >10 years to maturity, SE GVB 10Y8, and the T-bill with 3 months to maturity, SSVX 3M6. OP =the first difference in the log of the OPEC basket of oil price. In the study by Gervais et al. (2001) they construct volume groups based on the past 50 days, which will also be use in this study. The criteria used to determine a high and low volume day is consequently today’s volume evaluated on the past 50 days MA (out-ofsample). If the volume is above (below) one standard deviation, assuming normality, of the 50 day MA it results in a high (low) volume day. Otherwise it is determined a normal volume day.9 Dummy for volume group G1 (high volume group): 1 if > one stdev of the 50 day MA of volume, otherwise 0. Dummy for volume group G3 (low volume group): 1 if < one stdev of the 50 day MA of volume, otherwise 0. 7 SSVX 1M and SSVX 3M, a Swedish Treasury Bill with 1 and 3 months maturity, respectively. SE GVB 10 Y, a Swedish Government Bond with 10 years maturity 9 In the study by Gervais et al., (2001) they determines their formation periods using the top/bottom 10% of the daily volumes over the whole trading interval. A 10% limit is determined to be too strict and would further capture too few observations. Based on the fact that only firm and macro announcements, contributing to high volume, occurs around 6-12 times per year. Where the volume is expected to be high before and after the announcement for a few days. This alone would results in >10% of the total trading days in any given year. 8 15 G#TB = Dummy for group G# times the TB G#TS = Dummy for group G# times the TS G#OP = Dummy for group G# times the OP where # = 1 or 3. 3.4 Regression model10 The standard Ordinary Least Square method is used where the regression model is presented below. 𝑅𝑡 = 𝛼 + 𝑇𝐵𝑡 + 𝑇𝑆𝑡 + 𝑂𝑃𝑡 + 𝐺1𝐷𝑢𝑚𝑚𝑦 ∗ (𝑇𝐵𝑡 + 𝑇𝑆𝑡 + 𝑂𝑃𝑡 ) + 𝐺3𝐷𝑢𝑚𝑚𝑦 ∗ (𝑇𝐵𝑡 + 𝑇𝑆𝑡 + 𝑂𝑃𝑡 ) + 𝜀𝑡 𝜀𝑡 ~𝑁(0, 𝜎𝑡2 ) Newey and West (1987) heteroskedasticity-consistent standard errors are used, consistent with the general perception about the characteristic of financial data. This approach is used to handle issues with autocorrelation and heterskedasticity common in financial data. Graphical tests are performed along with standard econometric tests; heteroskedasticity tests, (White, 1984) autocorrelation tests Breusch-Godfrey Serial correlation LM test and Durbin-Watson, (Godfrey, 1978, 1981; Durbin and Watson, 1951), unit root test (Dickey and Fuller, 1979), and normality11 tests (Jarque and Bera, 1980) for the relevant time-series used in the study. However, no peculiar finding should be found since the explanatory variables are constructed using standard procedure for financial data. Consequently, following this procedure for the input data the risk of jeopardizing the reliability of the results are reduced, and the coefficient estimates should be the best linear unbiased estimators (BLUE). 10 Brooks, (2008) Even though it is empirically concluded several decades ago the returns are not normally distributed (Mandelbrot, 1997), they are assumed to be normally distributed in this study. This assumption is consistent with most academic research on financial data. Given the large sample size the impact of nonnormality is further limited. 11 16 3.5 Hypothesis discussion Group 1, G1, consists of days with high volume, in general generated by large institutional investors. Informed and sophisticated they are expected to signal information which reflects the conditions in the market. Given their size and experience they are primarily expected to act when they have an information advantage and/or when new information is released. In doing so they are likely to move the market. Consequently, the activity from the high volume days is expected to be related to the explanatory variables. (Cohen et al., 2002; Daniel et al., 1997; Holden and Subrahmanyam, 1992; Nofsinger and Sias, 1999, Shu, 2010) There is also the possibility that the explanatory variables are not a good reflection. This as the high volume days are expected to be correlated to firm and macro specific news announcements limiting the power of the selected explanatory variables, shifting it to the actual news event and other variables not accounted for. Group 2, G2 consisting of days with normal volume. Made up of a mix of institutional and individual investors continuously active in the market the relation with the explanatory variables is expected to be consistent with that observed in G1. The variables in normal days are further expected to demonstrate a higher relation to the explanatory variables than for the other two groups. Group 3, G3, consists of days with low volume. The expectation from days with low activity is that they are poorly related to the explanatory variables as the investors’ uncertainty is higher during low volume when they are awaiting new information. Expressed differently, little consensus which one can act upon is available. (Lee and Swaminathan, 1998) Second, during days with low volume there is an expected greater share of uniformed investors active in the market. (Stickel and Verrecchia, 1994) This group of investors does suggestively not rely as much on the explanatory variables in their decision making as the other more sophisticated groups. (Cohen et al., 2002; Daniel et al., 1997; Holden and Subrahmanyam, 1992; Nofsinger and Sias, 1999) Hence, the potential impact it could have on the return, positive or negative, is undetermined. This further provides an explanation to why the return could be poorly associated to the explanatory variables during low volume days. 17 3.5.1 Hypotheses table Relation 1M T-Bill Term spread Oil price G1 - + +/- G2 - + +/- G3 +/- +/- +/- In the table above a summary of the expected relation among volume activity and variables are presented. To support the results and the study firms will be sorted by sector, following the global industry classification standard, GICS, and by turnover. The turnover will be calculated as the average price times the average volume. The firms will then be divided into a high and a low turnover group. This will be conducted so that the sizes of the two groups are not too skewed while still reflecting an intuitive break of the turnover. 18 3.6 Methodological difficulties Stocks lacking a continuous historical data records will be excluded from the study. This was the case for Securitas and Nokia. If data is missing for the explanatory variables the most recent value were used. The start day for the regression model is 1992-01-02 and the final day is 2011-04-29. A list of all firms and sectors can be found in table 10 in the appendix. Since this study relies upon parts rather than the whole of other researchers work, with a different approach and purpose the assumptions of using a 50 day moving average and one standard deviation as filtering criteria was carefully investigated. Test results revealed that using longer moving averages, 100 and 150 days, would cause limited changes on the number of observations in the normal volume group. However, it contributed to an asymmetric size of each extreme volume group. Hence, a 50 day moving average was determined good. However, tests of using one standard deviation to filter out high and low volume days was not satisfactory to create symmetric sized groups consisting of approximately 16 % of the observations each. This due to the non-normality characteristics of the volume data. Therefore, using trial and error it was concluded that 0,82 standard deviations was best. This assured that each group contained a satisfactory share of the observations approximately corresponding to the characteristics of a normal distribution.12 12 1,00 stdev, 0,75 stdev, 0,80 stdev, 0,85 stdev and 0,84 stdev has been tested to reach consensus for the complete sample of stocks. Note also that it was optimized to fit all firms in the sample and and individual optimization could improve the results. 19 4 Empirical Results In this fourth chapter I present the empirical findings from the conducted study. All results are revealed in graphs and tables accompanied by brief comments. In reflecting over the results there will be a focus on OMXS30, a summary of the index constituents and brief comments from a turnover/sector comparison. 4.1 The OMXS30 index 4.1.1 Descriptive statistics of the volume and return Table 1 –Volume characteristics for different volume groups of OMXS30 Descriptive statistics for the G2 group is presented to give the reader an idea about the characteristics during a day with normal volume. The volume distribution in all groups are demonstrating a non-normal distribution with rather different characteristics. We note that each volume group G1 (high volume) and G3 (low volume) contains more observations than desirable using the same specification as for individual firms, 0.82 standard deviations. This contributes to G2 containing too few observation and the extreme groups being asymmetric in size. (After using 0.87 standard deviations as filtering characteristic more symmetric size volume groups was obtained. The overall impact of the change was on the other hand very small and the relations between the groups persisted.) 20 Table 2 – Return characteristics for different volume groups for OMXS30 The return characteristic does exhibit considerably normal distributional characteristics for all volume groups. The highest average return occurs in the normal volume days, 0.065%. For the high volume days the return is negative (-0.042%), while for low volume days it is positive (0.037%) on average. It further demonstrates that large extreme returns do, not unexpectedly, occur under high volume. Please see Appendix, Table 23 for a graph of the high and low volume day’s market in the price graph together with the daily returns. The graph reveals that there is a large tendency to volume clustering, where low volume in general characterize short-term consolidation periods. Table 3 – Correlation matrix (The SEKUSD time-series does not start until 1994-01-03). Please see appendix, table 24 for a graph of the explanatory variables. Above it is noted that the variables demonstrate very low cross-correlation within the sample period. The highest relation is observed for the oil price and SEKUSD, -0.17.13 13 The author considered it important to investigate the relation between USDSEK and oil price to develop an idea about the possibility of USDSEK price changes being incorporated in the oil price. 21 4.1.2 Results from the regression model, OMXS30 In discussing the results for the normal volume activity this is reflected by the TB, TS, and OP coefficients. Table 4 – Regression results from the OMXS30 index14 The regression results above reveals that the term spread, TS, and the oil price on high volume days, G1OP, are positive and significant. It is moreover concluded that the term spread by itself account for ~54% of the mean return for the OMXS30 index under normal volume, which is noteworthy. However, when also accounting for days with high volume it drops down to ~38%. (A regression was also run when 0,87 standard deviations was used as the filtering criteria. The above presented coefficient estimates was still a good approximation where the p-value for OP, G1TB and G1TS approached, but never reached, significance at the 10% level.) 14 A regression of only the TB, TS and OP without a volume filter resulted in TS and OP being significant. 22 4.2 The OMXS30 index constituents 4.2.1 Descriptive statistics of the volume and return Table 5 –Summary of the volume within different volume groups for all firms This table illustrate the turnover statistics for the different volume groups. It also presents the distributional characteristics for each volume group and the spread among all the firms. Example, in volume group G1 in the Average (turnover) row one see the distribution characteristics for the turnover among all firms. On the Obs./total row the distributional characteristics for the ratio of observations amongst all firms are presented. Example, in the dark grey cell above, representing the high volume group, we see that for the ratio of observations as part of the total for the median firm in the sample is 16.25 %. (Average (turnover) is calculated as the average volume * the average price for 15 each group.) In the table above we note the 50 day moving average of the volume with 0,82 standard deviations as filtering criteria for each group is a good specification. The percentage share of the average and the median volume is seemingly equal, 16.06% for the high volume group and 15.79% for the low volume group. This even though there is a considerable skewness and kurtosis within the high volume group (G1) and low volume group (G3). 15 See table 11-15 in appendix for details of volume distribution for each individual firm, sorted either by sector or turnover. 23 Table 6 –Summary of the return within different volume groups for all firms On the horizontal axis the return for all firms using different statistical features are presented. Example, in the Complete Period section directly above, the average in the left column (y-led, dark grey) is the average return for a firm. Then go right, (x-led, light grey) to the 50th percentile and you see the median average return among all firms, 0.044%. If you look further to the rights, in the stdev column (x-led, lightest grey) you see the standard deviation of the average returns for all firms in the group Complete Period, 0.026%16 The return characteristics corresponding to each volume group, found in the table above, illustrate interesting results. The average return and standard deviation (stdev) demonstrates a decaying pattern for the average, 25th, 50th, and 75th percentile separation across volume groups (look vertical/top-down). These figures are higher in high volume group and lower in the low volume group. The difference is considerable where the average (standard deviation) is 0.250% (3,833%) for the high volume days, 0.017% 16 See table 11-15 in appendix for details of volume distribution for each individual firm, sorted either by sector or turnover. 24 (2.150%) for the medium volume days, and -0.065% (1.581%) for the low volume days while over the complete period it is 0.043% (2.436%). It is also noted that the skewness and kurtosis for medium and low volume is in between -0.18 – 0.31 and 2.21 – 4.65 respectively. Hence, it is fairly normally distributed around its average. While in the high volume group the kurtosis is considerably higher. (The overall pattern also persists after removing two outlier firms, ABB and SKF). 4.2.2 Descriptive statistics for OMXS30 constituents sorted by turnover and sector If the reader is interest in descriptive statistics for each individual firm in the OMXS30, tables can be found in Appendix sorted by turnover and sector (table 11-15). 4.2.3 Results from regression model, OMXS30 index constituents Table 7 – Summary of the regression results for ONLY significant variables among all firms In the table above the average coefficient values for the significant variables accompanied by the average p-value are presented. The standard deviation is calculated for the coefficient values and the p-values. Example, for the TS variable it is significant for 15 out of the 27 firms. Among these 15 firms the average coefficient value is 0.0012, with a standard deviation of 0.0006. As regards of the average p-value for the significant variables it is 0.03299 with a standard deviation of 0.02704. At last we note that the TS impact in relation to the constant, C, is -35.83%. If the reader is interested in the coefficient and p-values for each individual firm these can be found in appendix in table 16-17 sorted by sector and in table 18-19 sorted by turnover. In the table above we conclude that the constant is negative and significant for 26 % of all the firms in the sample. The oil price (OP) is the variable demonstrating the best explanatory characteristics. It is significant for 81 % of all firms on normal days and is moreover significant for 22% of the firms among the high volume group and 15% among the low volume group. For those firms which the variables are significant it is interesting to note that 11.13%17 of the oil price return is transferred over to the average stock return 17 9.28% excluding Lundin Petroleum 25 on normal volume days, and as much as 47.73% (36.60%18 + 11.13%) on high volume days. However, the overall impact transferred to a firm’s average return is just a few percent, at most. It is noted that for the normal and high volume days the oil price coefficient is positive. Important to stress is that the sign of the explanatory variable G1OP is negative on average, while for the low volume group G3OP it is positive. (Please see Appendix, Table 22) The second most significant variable, for 56 % of the firms, is the term spread (TS) under normal volume. Concerning the high volume group the relation is also positive and significant for 48 % of the firms. In the low volume days the relation is negative and significant for 30 % of the firms. Noteworthy is that the TS across the three volume groups is the single most important factor in explaining the return across all firms. Shifting focus to the 1 month T-bill the explanatory power is the lowest. It is negative and significant for 33 % of the firms in the normal volume group, positive and significant for 22 % of the firms in the high volume group, and positive and significant for 41 % of the firms in the low volume group. However, the impact on the return is very small. 4.2.4 Results from regression model results, OMXS30 index constituents sorted by turnover and sector In the Appendix results for individual firms are presented sorted by turnover and sector. According to statistical test no difference was found for sectors versus the whole sample. In the table illustrating the sector group separation, each firm’s regression coefficients are presented accompanied by its impact in relation to the stocks mean return. It reveals that the term spread indeed can have a very high impact on the return for many firms. (Please see Appendix, Table 16-17,21) When firms are sorted by turnover more interesting results are found. It is statistically concluded that the explanatory variables are better explaining high turnover stocks than 18 25.24% excluding Lundin Petroleum 26 low turnover stocks. The average number of explanatory variables for the high turnover stocks is 4.07 versus an average of 3.38 for low turnover stocks, with a standard deviation of 1.60. For the high turnover stocks the variables more significant than the low turnover stocks are the constant (C), the term spread (TS and G3TS), the 1 M T-Bill (G1TB), and the oil price (OP). The opposite relation is found for the term spread (G1TS), which is significantly more frequent among the low turnover stocks. In the table presenting firms sorted by turnover, each firm’s regression coefficients is presented in relation to the constant. (Please see Appendix, Table 18-19, 20) 4.3 Concluding remarks and main results It was found that the term spread and oil price was the most important variables to explain the index and stock return. The term spread demonstrated the most significant impact on the return. For individual firms it was found that high turnover stocks return was better explained using these variables than low turnover stocks. 27 5 Analysis and discussion This chapter begins with a discussion of OMXS30 and its constituent firms. It is followed by a discussion of the explanatory variables and the sector/turnover comparison. In analyzing the return characteristics and the volatility for each volume group the results are consistent with what was expected according to Lee and Swaminathan (1998). They argued, and the results illustrate it, that high volume days are subject of greater risk and further demonstrate a higher return volatility, while the opposite holds true for the low volume days. (Please see appendix, table 11-15) Shu, (2010) found that stocks with higher fraction of institutional trading volume outperform stocks with lower fraction of institutional trading volume. Although this study have not performed a explicit separation between institutional and individual investors consistent with Shu (2010), the return characteristics demonstrate that under high volume the return is significantly larger than for the low volume group across firms. As argued previously it is not unlikely that these group formations would be a good representation of institutional and individual investors respectively. Interestingly it was noted that on high volume days for the index the average return was negative contrary to individual stocks. This is likely due to Ericsson, Sandvik and ABB which throughout the sample period demonstrate a negative return in the high volume period and further has a high weighting in the OMXS30 index. This illustrates the importance of considering individual analysis in relation to index research. Table 8 – Hypothesis table, OMXS30 index Relation and significance 1M T-Bill, TB Term spread, TS Oil price, OP G1, High. - - + and significant G2, Normal - + and significant + G3, Low + + + The above table illustrates the sign of the coefficients and if it was significant for the OMXS30 index. In relation to formulated hypothesis the results for the OMXS30 index illustrate fairly consistent results. As hypothesised, the impact on low volume days was difficult to determine but it turned out to be positive for all variables. For the normal and high volume days the TB had a negative impact which was expected. The OP had a positive 28 impact which was close to being significant for both volume groups. Regarding the TS the results were mixed. The regression results for the OMXS30 index revealed limited significant variables, especially for the extreme volume groups. This is interesting since the weighting of stocks in the OMXS30 do contain a high share of high turnover stocks, a group that demonstrated more significant variables than that for the low turnover group. Consequently, a low explanatory power for the exogenous variables during high and low volume groups for the OMXS30 index could be due to the index being subject of changes more related to, and reflecting, other macroeconomic figures. Figures such as the GDP, industrial production, export and import, money growth, and global economic figures among others. Table 9 – Hypothesis table, OMXS30 index constituents, relation and significance Relation 1M T-Bill, TB Term spread, TS Oil price, OP G1, High + + + (12 firms have + sign) (25 firms have + sign) (22 firms have + sign) - + + (16 firms have – sign) (25 firms have + sign) (26 firms have + sign) + (20) - (24) - (18) (20 firms have + sign) (24 firms have - sign) (18 firms have - sign) 1M T-Bill, TB Term spread, TS Oil price, OP G1, High 3 firms, a significant + sign 13 firms, a significant + sign 6 firms, a significant + sign volume 3 firms, a significant - sign 0 firms, a significant - sign 0 firms, a significant - sign G2, Normal 3 firms, a significant + sign 15 firms, a significant + sign 21 firms, a significant + sign volume 6 firms, a significant - sign 0 firms, a significant - sign 1 firm, a significant - sign 10 firms, a significant + sign 0 firms, a significant + sign 2 firms, a significant + sign 8 firms, a significant - sign 2 firms, a significant - sign volume G2, Normal volume G3, Low volume Significance G3, Low volume 1 firm, a significant - sign The relation hypothesis table above illustrates the sign, on average, for all firms in the study. Under each sign, in parenthesis, the number of firms which demonstrate that sign is presented. Example, in the normal period there is on average a negative relation between the 1 month T-bill and the stock return, and 16 firms out of 27 demonstrate a negative sign. In the significance hypothesis table the number of firms with significance and the relation is presented. If the reader is interested in the relation and significance for an individual firm please see appendix in table 16-17 sorted by sector and table 18-19 sorted by turnover. 29 In analyzing the results from the OMXS30 firm constituents several findings were observed, see table 9. As indicated in the theory section the explanatory variables relation in the low volume days was expected to show low significance and mixed results. This is what the results illustrate concerning the OMXS30 index and across firms. For the three variables in the low volume group only one is consistent with what was observed for the index, which is the TB. In the normal volume group the TB is negative on average and that for the majority of the firms. In contrast to the hypothesis and what was found for the OMXS30 index there is a positive relation, on average, for the TB in the high volume group. However, it is among the minority of the firms that this positive relation is demonstrated. The results for the TS and OP in the normal and high volume groups are clearer and demonstrate a positive sign for almost all firms, consistent with the hypothesis. As regards of the negative relation among firms in the low volume group for TS and OP this likely reflects that little information is available in the market. And with little information the uncertainty increases and a negative or undetermined relation would be expected as demonstrated. Concerning the significance for individual firms it is clear that the TS is indeed a good and consistent explanatory variable under the three volume groups, followed by the OP. In discussing the explanatory variables, starting with the TB, it was in line with the expectations demonstrating a negative relation for the OMXS30 index (table 8) and that also for the majority of the firms under worth mentioning volume. It is interesting to note however that under low volume the relation is positive and significant for 10 firms. Important to remember is that the impact on the return for the TB was low. This is not unlikely as the effect of the financing costs on a firm’s total profit is small in most cases and of less importance to the stock return, in comparison to many other factors. Moreover, most of the firms within the OMXS30 have a significant cross-border operation. This enables these firms to raise capital on the international market and would possibly be more subject of international interest rates movements. In large the results for the short term interest rate variable were consistent with previous researchers, such as Avramov and Chordia (2006) and Perez-Quiros and Timmermann (2000) among other. 30 The 1 month T-Bill demonstrated a negative relation with the return most likely reflecting higher financing costs. As regards of the term spread a consistent relation with previous research was found, with the exception of high volume days for the OMXS30 index, and the underlying reason for this is hard to determine. For Swedish stocks and its major index the term spread was the most important variable across all volume groups and firms. It is concluded a good measure in general and not only a good predictor of recessions as suggested by Estrella and Mishkin (1998) and bull and bear markets as suggested by Chen (2009). These results are contrary to Rapach et al., (2005) which found limiting evidence for the terms spread as an explanatory variable. Important to note is that the impact of the term spread in relation to the average return over the sample period, especially on days with normal volume is extreme for several firms (please see appendix, table 16-17). That the term spread should account for several hundred percent of the average return is spurious. The conclusion from this is that additional interest rate variables seem to be justified as they could help narrow down the relation between the observed return and that of other explanatory variables. The peculiar finding, in contrast to previous research such as Driesprong et al. (2008) and Nandha and Faff (2008) suggesting a negative relation with the oil price, is that it demonstrates a clear positive relation with the return for all firms but one. And on normal days this relation is significant for four fifths of the firms. This seems to suggest that Swedish firms benefit from higher oil prices. This could have several explanations. As mentioned previously one is that higher oil prices reflect a booming economy which would be positive for most firms. Another explanation is that the Swedish economy and its firms, although representing several energy intensive sectors, are less dependent on fossil fuels usage than perhaps many other firms and countries. Hence, with less dependence on oil as an input variable they are compensated when oil prices increases. This either from adjusting their own prices resulting in an improved profit margin, or an increased competitiveness increasing the demand for their goods and services. Using oil as a proxy for inflation seems not unlikely given its importance in society today. In doing 31 so the expected relation suggested by the fisher theory, predicting that stocks provide a hedge against inflation, holds. If the average sign of the explanatory variable (the input data) is incorporated in the analysis, found in table 22 in appendix, one note that the total effect is that under high and low volume the impact form the oil price impact is on average negative for the stock return, while under normal volume the impact is positive on average for the stock return. Concerning the sector and high versus low turnover stock comparison it was hard to reveal any differences for sectors. First, most of the different sector groups were small with only a few firms within each, plus within each sector the sub-industry each firm operated in were considerable different. A more intuitive comparison could have been services providers versus goods producers. This could capture some firms’ high physical capital costs and energy intensive production contrary to other firms low capital costs and high human capital costs. The separation of stocks into a high and low turnover group revealed more interesting findings. The volume for high turnover stocks contained information which was to a higher degree related to the explanatory variables than the lower turnover stocks. The average return was further significantly higher for high turnover stocks than low. This in line with what several other researchers have suggested. (Copeland, 1976; Easley and O´hara, 1992, Bessembinder and Segin, 1993; Gervais et al., 2001) However, as regards of the standard deviation of the return it was significantly lower for high turnover stock than for low turnover stocks. One explanation for this is that the high turnover stocks are more governed by institutional investors and finance journalists, nationally and internationally. This reduces information asymmetries and the high turnover stocks would possibly face lesser degree of news announcement surprises than lower turnover stocks might. Algorithm trading, more implemented on high turnover stocks, could also play a role in which the computerized trading contribute to lower volatility as they seize mispricing faster. In summary, Lamoureax and Lastrapes (1990) presenting that trading volume is a good proxy for the arrival of new information, Stickel and Verrecchia (1994) arguing that as volume increases the likelihood that the price change is information drive, and finally Clark (1973) conclusion that there is a relation between trading volume and volatility are supported by the results. 32 6 Conclusion In this final chapter I present my conclusions from the performed study. I reflect on the research and finally I give suggestions to further research around some of the fields this study have touched upon. A considerable amount of research was analyzed to find consistent and intuitive explanatory variables. Three was found, the 1 M T-Bill, the term spread - 10 Y Treasury bond versus a 3 M T-Bill -, and the oil price. In trying to reveal and better understand the complexity of the financial markets these three variables was used together with a volume filter for high, normal and low volume days. The results from using this methodology have revealed that accounting for trade volume is important in trying to explain the return. The chosen explanatory variables do indeed explain Swedish index and stock returns as suggested by previous research. Moreover, filtering for the volume provides additional insights of when the explanatory variables are useful. It provides insights on the relation; sign and size of the impacts, which varied significantly across different volume activity across firms. The results revealed a significant difference between high and low turnover stocks. The most reliable and consistent variables were the oil price followed by the term spread, both demonstrating a positive relation with the return. 6.1 Criticism of research The results rest upon an in-sample study revealing the sign and size an explanatory variable has in relation to the stock return. However, it does not say anything about whether a change in the oil price or the interest rate variables comes pre or post a stock price movement. Hence, it is important to further investigate the causality between the stock return and the explanatory factors for the Swedish case. It is also important to be aware of that a lot of economic research conducted using macroeconomic data use lower frequent data such as weekly or monthly observations. This data does not include as much noise as the daily observations does used in this study. 33 Although, an extensive literature review was conducted to find appropriate variables general for all firms more could have been chosen and rather let the regression decide which are useful and which are not. However, using the chosen volume filter three times as much information must be interpreted. That and given that hardly any other research has been found using a similar methodology it was necessary to impose some restrictions to be able to target the possible relations and reach a conclusion if the methodology chosen to work with was promising or not. 6.2 Further studies The results and the methodology are indeed promising and a bigger study with large cap, medium cap and small cap stocks included would be interesting to conduct. This would make the sector and turnover group study more intuitive, and accurate. In such a study it would also be appropriate to include more explanatory variables trying to find a good model fit. It should further be interesting to study the trade records from higher frequent data. In doing so the focus would be to separate the trade activity relating to the institutional investors from that of the individual investors. In this and above mentioned scenario it would moreover be important to conduct out-of-sample test to reveal if the results are useful in practice. An extensive amount of data have been researched and analyzed. The results are encouraging and reveal a lot of interesting relations, some of which have been discussed around in the text while many more can be revealed if the reader study the tables and graphs more closely. The filtering methodology show potential and deserve further research. 34 7 References Articles and Books Abbondante, P., 2010, Trading volume and Stock indicies: A test of Technical Analysis, American Journal of Economics and Business Administration 2, p. 287-292 Adler, M., and Dumas, B., 1983, International portfolio choice and corporate finance: A synthesis, Journal of Finance 38, p. 925-984 Ang A., and Bekaert, G., 2001, Stock return predictability: Is it there?, National Bureau of Economic Research, Working Paper No. 8207 Apergis N., and Miller, S.M., 2008, Do structural oil-market shocks affect stock prices? Energy economics 4, p. 569-575 Avramov D., and Chordia, T., 2006, Predicting stock returns, Journal of Financial Economics 82, p. 387-415 Banz, R.W., 1981, The relationship between returns and market values of common stocks, Journal of financial economics 9, p. 3-18 Basu, S., 1983, The relationship between earnings yield, market value, and return for NYSE common stocks: Further evidence, Journal of Financial Economics 12, p. 129-156 Balvers R.J., Cosimano, T.F., and McDonald, B., (1990), Predicting Stock Returns in an Efficient Market, Journal of Finance 45, p. 1109-1128 Bessembinder H., and Segin,P.J., 1993, Price volatility, trading volume, and market depth: evidence from the futures market, Journal of Financial and Quantitative Analysis 28, p. 21-39 Bodie, Z., 1976, Common stocks as a hedge against inflation, Journal of Finance 3, p. 459-470 Barsky R.B:, and Kilian, L., 2004, Oil and the Macroeconomy since the 1970s, Journal of Economic Perspectives 18, p. 115-134 35 Black, F., 1976, Studies in stock price volatility change, Proceedings of the 1976 business meeting of the business and economic section, American Statistical Association, p. 177181 Bolleslev, T., 1986, Generalized Autoregressive Conditional Heteroskedasticity, Journal of Econometrics 31, p. 307-327 Brown, S.P.A., and Yücel, M.K., 1999, Oil prices and U.S. aggregate economic activity, Federal Reserve Bank of Dallas Economic Review 14, p. 16-53 Brown, S.P.A., and Yücel, M.K., 2002, Energy prices and aggregate economic activity: an interpretative survey, Quarterly Review of Economics and Finance 42, p. 193-208 Brooks, C., (2008), Introductory Econometrics for Finance, Cambridge University Press, Cambridge Bryman, A., Bell, E., (2003) “Business Research Methods”, Oxford University Press, Oxford Campbell, J. Y., 1987; Stock returns and the term structure, Journal of Financial Economics 18, p. 373-399 Campbell, J.Y., 1990, Measuring the persistence of expected returns, The American economic review 80, p. 43-47 Caruth, A.A, Hooker, M.A. and Oswald, A.J., 1998, Unemployment equilibria and input prices: theory and evidence from the United States, Review of Economics and Statistics 80, p. 621-628 Clark, P.K., 1973, A subordinated stochastic process model with finance variance for speculative prices, Econometrica 41, p. 135-155 Chan, L.K., Hamao, Y., and Lakonishok,J., 1991, Fundamentals and stock returns in Japan, Journal of Finance 46, p. 1739-1789 Chan, L.K.C., Karceski, J., and Lakonishok, J., 1998, The risk and Return from factors, The Journal of Financial and Quantitative Analysis 33, No 2, p. 159-188 36 Chang, K-L., 2009, Do macroeconomic variables have regime-dependent effects on stock return dynamics? Evidence from the Markov regime shifting model, Economic Modelling, 1283-1299 Chen, N-F., Roll, R., and Ross S.A., 1986, Economic forces and the stock market, Journal of Business 56, p. 383-403 Chen, N.F., (1991), Financial investment opportunities and the macroeconomy, Journal of Finance 46, p. 529-554 Chen, S-S., 2009, Predicting the bear stock market: Macroeconomic variables as leading indicators, Journal of Banking and Fiannce, 211-223 Chen, S-S., 2010, Do higher oil prices push the stock market into bear territory?, Energy Economics, 490-495 Claessens, S., Dasgupta, S., and Glen, J., 1995, The Cross-Section of Stock Returns, Evidence from Emerging Markets, Policy Research Working Paper, 1-20 Cohen, R.B., Gompers P.A., and Vuolteenaho, T., 2002, Who underreacts to cash-flow news? Evidence from trading between individuals and institutions, Journal of Financial Economics 66, 409-462 Conover, Jensen & Johnson, 1999, Monetary environments and international stock returns, Journal of Banking and Finance 23, p. 1357-1381 Copeland, T.E., 1976, A Model of assets trading under the assumption of sequential information arrival, Journal of Finance 31, p. 1149-1168 Chordia and Swaminathan (2000), Trading volume and Cross-Autocorrelations in stock returns, Journal of Finance 55, p. 913-935 Cutler, D, Poterba and Summers, 1989; International evidence on the predictability of stock returns, Working paper, Massachusetts Institute of Technology Davis S.J. and Haltiwanger J., 2001, Sectoral job creation and destruction responses to oil price changes, Journal of Monetary Economics 48, p. 465-512 Daniel, K., Grinblatt, M., Titman, S., Wermers, R., 1997, Measuring mutual fund performance with characteristic-based benchmarks, Journal of Finance 52, p. 1035-1058 37 Dickey D., and Fuller, W., 1979, Distribution of the estimators for autoregressive time series with unit root, Journal of American Statistical Association 74, p. 427-431 Driesprong G., Jacobsen, B., Maat, B., 2008, Striking oil: Another puzzle?, Journal of Financial Economics 89, p. 307-327 Dumas B., and Solnik B., 1995, The world price of foreign exchange risk, Journal of Finance 50, p. 445-479 Durbin, J., and Watson, G.S., 1951, Testing for Serial Correlation in Least Squares Regression, Biometrika 38, p. 159-71 Easley, D., and O’hara M., 1992, Time and process of security price adjustments, Journal of Finance 47, p. 577-605 El-Sharif, I., Brown, D., Bruce, B., Nixon, B., and Russel A., 2005, Evidence on the nature and extent of the relationship between oil prices and equity values in the UK, Energy economics 27, p. 819-830 Engle , R.F., 1982, Autoregressive conditional heteroskedasticity with estimates of the variance of United Kingdom inflation, Econometrica 50, p. 987-1007 Epps T.W., and Epps M.L., 1976, The stochastic dependence of security price changes and transaction volumes: implications for the mixture-of-distributions hypothesis, Econometrica 44, p. 305-321 Estrella, A., and Mishkin, F.S., 1998, Predicting U.S. recessions: Financial variables as leading indicators, Review of economics and statistics 80, p. 45-61 Faff and Brailsford, 1999; Oil price risk and the Australian stock market - some further results, Journal of Energy Finance and Development 4, p. 69-87 Fama, E., 1981, Stock returns, real activity, inflation and money, American Economic Review 71, p. 545-565 Fama, E., and French, K., (1992) "The Cross-Section of Expected Stock Returns”, Journal of Finance 47, p. 427-465 Fama, E.F., and Schwert, W., 1977, Asset returns and inflation, Journal of Financial Economics 5, p. 115-146 38 Fama, E., and French, K., 1989, Business conditions and expected returns on stocks and bonds, Journal of Financial Economics 25, p. 23-49 Ferson W., and Harvey, C., 1993, The risk and predictability of International Equity Returns, Review of Financial Studies 6, p. 527-566 Fleming, J., Kirby., Ostdiek, B., 2005, ARCH Effects and Trading Volume, Rice University and Clemson University Working Paper Flannery, M.J., and Protopapadakis, A.A., 2002, Macroeconomic factors do influence aggregate stock returns, Review of Finanical studies 15, p. 751-782 Feldstein M.S., and Eckstein, O., 1970, The fundamental determinants of the interest rate, Review of economics and Statistics 52, p. 363-376 Ferderer, J.P., 1996, Oil price volatility and the macroeconomy, Journal of Macroeconomics 18, p. 1-26 Fuhrer, J.C., 1995, The Phillips curve is alive and well, New England Economic Review of the Federal Reserve Bank of Boston March/April, p. 41-56 Gallant, A., Ronald, G.A., Rossi P.E., and Tauchen, G., 1992; Stock price and volume, Review of Financial Studies 14, p. 1-27 Gervais, S., Kaniel, R., and Mingelgrin, D., (2001), The High Volume Return Premium, Journal of Finance 56, p. 877-919 Geske, R., and Roll, R., 1983, The fiscal and monetary linkage between stock returns and inflation, Journal of Finance 38, p. 1-33 Gisser M., and Goodwin, T.H., 1986, Crude oil and the macroeconomy: tests of some popular notions, Journal of Money, Credit and Banking 18, p. 95-103 Glosten, L.R., Jagannathan, R.R., Runkle, D., (1993), On the relation between the expected value and the volatility of the nominal excess return on stocks, Journal of Finance 48, p. 1779-1801 Godfrey, L.J., 1978, Testing against general autoregressive and moving average error models when the regressors included lagged dependent variables, Econometrica 46, p. 1293-1302 39 Godfrey, L.J., 1981, On the invariance of the Lagrance multiplier test with respect to certain changes in the alternative hypothesis, Econometrica 49, p. 1443-1455 Gordon, R.J., 1997, The time-varying NAIRU and its implications for economic policy, Journal of Economic Perspectives 11, p. 11-32 Hamilton, J.D., 1983, Oil and the macroeconomy since World War II, Journal of Political Economy 91, p. 228-248 Hamilton, J.D., 2005, Oil and the macroeconomy, Working paper UCSD Hodrick, R.J., 1989, Dividend yields and expected stock returns: Alternative procedures for inference and measurement, Review of Financial Studies 5, p. 357-386 Holden C.W:, and Subrahmanyam, A., 1992, Long-lived private information and imperfect competition, Journal of Finance 47, p. 247-270 Hooker, M.A., 2002, Are oil shocks inflationary? Asymmetric and nonlinear specifications versus changes in regime, Journal of Money, Credit and Banking 34, p. 540-561 Huang, R.D., Masulis, R.W., Stoll, H.R., 1996, Energy shocks and financial markets, Journal of futures markets 16, p. 1-27 Jaffe, J., and Mandelker, G., 1976, The “Fisher Effect” for risky assets: An empirical investigation, The Journal of Finance 31, p. 447-458 Jarque, A., and Bera, A., 1980, Efficient test for normality, heteroskedasticity and serial dependence of regression residuals, Economic Letters 6, p. 255-259 Jones, C.M., and Kaul, G., 1996, Oil and the stock markets, Journal of Finance 51, p. 463-491 Karpoff, J.M., 1987, The relation between price changes and trading volume: A survey, Journal of Financial and Quantitative Analysis 22, 109-126 Keim, D., and Stambaugh, R.F., 1986, Predicting returns in the stock and bond markets, Journal of Financial Economics 17, p. 357-390 40 Kilian, L., and Park, C., 2009, The impact of oil price shocks on the U.S. stock market, , International economic review 50, p. 1267-1287 Keane, M.P., and Prasad, E.S., 1996, The employment and wage effects of oil price changes: a sectoral analysis, Review of Economics and Statistics 78, p. 389-400 Lamont, O., 2001, Economic tracking portfolios, Journal of Econometrics 105, p. 161184 Lardic, S., and Mignon, V., 2006, The impact of oil prices on GDP in European Countries: an empirical investigation based on asymmetric cointegration, Energy Policy 34, p. 3910-3915 Lardic, S., and Mignon, V., 2008, Oil prices and economic activity: an asymmetric cointegration approach, Energy enomics 30, p. 847-855 Lee, C.M.C., and Swaminathan, B. 1998, Price momentum and trading volume, Johnson Graduate School of Management Cornell University LeBlanc, M. and Chinn M.D., 2004, Do high oil prices presage inflation? The evidence from G5 countries, Business Economics 34, p. 38-48 Lescaroux, F., and Mignon, V., 2008, On the influence of oil prices on economic activity and other macroeconomic and financial variables, OPEC energy Review December 2008 Litzenberger R.H., and Ramaswamy, K., 1982, The Effects of Dividends on Common Stock Prices Tax Effects or Information Effect?, The Journal of Finance 37, p. 429-443 Lewellen, J., 2004, Predicting returns with financial ratios, Journal of Financial Economics 74, No 2, p. 209-235 Llorente, G., and Michaely, R., Gideon, S., and Wang, J., 2002, Dynamic volume return relation of individual stocks, Review of Financial studies 15, p. 1005-1139 Mandelbrot, B.S., 1963, The variation of certain speculative prices, Journal of Business, 36, p. 394-419 Mandelbrot, B.B., and van Ness, J.W., 1968, Fractional Brownian motion, fractional noises and applications, parts 1,2,3, SIAM Review 10, p. 422-437 41 Mandelbrot, B.B., 1997, Fractals and Scaling in Finance, Springer-Verlag, New York Berlin Heidelberg Michael S., and Starks L.T., (1988), An Empirical Analysis of the Stock Price-Volume Relationship, Journal of Banking and Finance 12, p. 31-41 Mork, K.A., 1989, Oil and the macroeconomy when prices goes up and down: an extension of Hamilton’s results, Journal of Political Economy 97, p. 740-744 Mork, K.A., Olsen, O., and Mysen, H.T., 1994, Macroeconomic responses to oil price increases and decreases in seven OECD countries, The Energy Journal 15, p. 19-35 Mory, J.F., 1993, Oil prices and economic activity: is the relationship symmetric? The Energy Journal 14, p. 151-161 Mussa, M., 2000, The impact of higher oil prices on the global economy, International Monetary Fund 2000, 8 december, http://www.imf.org/external/pubs/ft/oil/2000/oilrep.PDF Nandha, M., and Faff, R., 2008, Does oil move equity prices? A global view, Energy Economics 30, 986-997 Nelson, D. B. (1991), Conditional Heteroskedasticity in Asset Returns: A new Approach, Econometrica 59, p. 347-370 Newey, W.K., and West, K.D., (1987), A simple, Positive Semi-Definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix, Econometrica 55, p. 703-708 Nelson, C.R., 1976, Inflation and rates of return on common stock, Journal of Finance 31, p. 471-483 Nofsinger J.R. and Sias, R.W., 1999, Herding and feedback trading by institutional and individual investors, Journal of Finance 54, p. 2263-2295 Park, J., and Ratti, R.A., 2008, Oil price shocks and stock market in the U.S. and 13 European countries, Energy Economics 30, p. 2587-2608 42 Patro D.K., Wald, J.K., and Wu, Y., 2002, The impact of macroecomnomic and FInanical variables on market risk: evidence form international equity returns, European Financial Management 8, p. 421-477 Pearce D.K., and Roley, V.V., 1983, The reaction of stock prices to unanticipated changes in money: A note, Journal of Finance 38, p. 1323-1333 Pearce D.K., and Roley, V.V., 1985, Stock price and economic news, Journal of Business 58, .p 49-67 Pesaran, M.H., and Timmermann, A., 1995, Predictability of stock returns: Robustness and economic significance, Journal of Finance 50, p. 1201-1228 Perez-Quiros, G., and Timmermann A., 2000, Firm size and cyclical variations in stock, Journal of Finance 55, p. 1229-1262 Pontiff and Schall, 1998, Book-to-market ratios as predictors of stock return, Journal of Financial Economics 49, p. 141-160 Poon, S-H., Granger, C., 2003, Forecasting volatility in financial markets: A review, Journal of Economic Literature 41, p. 478-539 Roll, R., 1992, Industrial structure and the comparative behavior of international stock market indexes, Journal of Finance 47, p. 3-42 Rapach, D. E., Wohar, M. E., and Rangvid, J., 2005, Macro variables and international stock return predictability, International journal of Forecasting, 137-166 Sadorsky, P., 1999, Oil price shocks and stock market activity, Energy Economics 21, p. 449-469 Schwert, W.G., 1989, Why does stock market volatility change over time?, Journal of Finance 52, p. 35-55 Shu, T., 2010, Trader composition and the cross-section of stock returns, The University of Georgia Stickel S.E., and Verrecchia R.E., 1994, Evidence that trading volume sustains stock price changes, Financial Analysts Journal, November-December, p. 57-67 43 Saunders, M., Lewis, P., Thornhill, A., (2003) Research Methods for Business Studies 3rd ed”, Pearson Education Limited, EssexSimon et al. (2001 Tauchen G.E., and Pitts M., 1983, The price variability-volume relationship on speculative markets, Econometrica 51, p. 485-505 Whitlaw, R.F., (1994), Time variations and covariations in the expectation and volatility of stock returns, Journal of Finance 49, p. 515-541 Databases: Datastream Advance 5.0, Thomson Financial Limited Internet NASDAQ OMX NORDIC, http://www.nasdaqomxnordic.com/nordic/Nordic.aspx Riksbanken, http://www.riksbank.se/templates/SectionStart.aspx?id=8720 44 Appendix Table 10 –Firms in the study with corresponding sector, industry group and sub-industry (GICS) The industries represented by the OMXS30 index constituents are diverse. Many firms operate with the whole world as their market, along with is large size it is also possible that an international rate, rather than the Swedish, should be used. Moreover, it is also highly likely that the USD and EUR exchange rate could have a significant impact. Of greater weight for the following comments and notes on the sector comparison is the diversity. Hence, a focus on the financial and industrial firms will be keep as these sectors demonstrate most similarities within their sub-industry group. 45 Table 11 – Descriptive statistics of firms sorted by turnover (1 of 2) For comments see under Descriptive statistics of firms sorted by turnover (2 of 2). 46 Table 12 – Descriptive statistics of firms sorted by turnover (2 of 2) Comments to Descriptive statistics of firms sorted by turnover 1-2. One can note that the skewness (skew) is largest for group G1 followed by G2 and G3, the same relation holds for the kurtosis (kurt). The average return for G1 is 0,25 %, being negative for 3 out of 27 firms. The average return for G2 is 0,02 %, and is positive for 19 out of 27 firms. The average return for G3 is -0,06%, being negative for 22 out of 27 firms. The standard deviation of the return (third column from the right) is largest in G1 and is gradually decaying for all firms to being the lowest for G3, the standard deviation for group G1 is larger than that for the complete period among all firms. The turnover for G1 is 3,5 to 5 times larger than that found for G3. 47 Table 13 – Descriptive statistics of firms sorted by sector (1 of 3) The financial companies demonstrate a large spread in turnover between themselves. Each firm has 17-19 % of total trade observations in G1 and G3 demonstrating considerably symmetric groups individually and cross-sectional. The average return is positive in group G1 and G2 and negative in G3. The exception is Swedbank which demonstrate a negative return on average in G2 and positive in G3. 48 Table 14 – Descriptive statistics of firms sorted by sector (2 of 3) From the industrial companies we note that the volume groups are not as symmetric as for the financial firms, there is also a larger spread cross-sectional where some groups having to many and other too few observations compared to what was sought after. The average return is positive (negative) in the complete period, G1, and G2 (G3). 49 Table 15 – Descriptive statistics of firms sorted by sector (3 of 3) Companies in the materials sector is, just as with the industrials, operating in very different sub-industries. The size of each volume could be better adjusted with individual determined standard deviations rather than using 0,82 which was optimized for the complete sample. The average is positive for the complete period and G1. Whilst for G2 and G3 the sign of the returns are mixed. 50 Table 16 – Results from regression model for firms sorted by sector, 1 of 2 (read together with table 17 and 21) 51 The single energy stock in the sample is the one demonstrating best features. For the health care sector it is particular the TB, OP, and G3TS that have good explanatory power, which could relate to high R&D costs and a high dependence on energy. However, the sign for the TB and OP is opposite each other for the two firms. For financial firms TS, OP, and G3TB show highest accuracy. Moreover, if one looks beyond a p-value of <0.10 versus 0.15-0.20 the TS emerges as unquestionably the one with most explanatory power. 52 Table 17 – Results from regression model for firms sorted by sector, 2 of 2 (read together with table 16 and 21) Moderately consistent results within sectors are found. The industrials sector demonstrating most, where OP, TS, G1TS, and G3TS have best explanatory power. 53 Table 18 – Results from regression model for firms for high turnover, 1 of 2 (read together with table 19 and 20) Above we note that the term spread (10 of 14) and oil price (12 of 14) is most significant variables for the high turnover group (>200,000,000). We can also note that the G3TB, G3TS, and G3OP is more significant than they are for G1TB, G1TS, and G1OP for firms > 400,000,000. Could higher monitoring explain this? 54 Table 19 – Results from regression model for firms for low turnover, 2 of 2 (read together with table 18 and 20) For the low turnover group (<200,000,000) we note that it is in particular the oil price (10 of 13) and high volume term spread, G1TS, (10 of 13) that demonstrate significant variables. The interesting finding here is that the term spread G1TS rather than TS is most significant for the low turnover group. Moreover, in total there are more significant variables for the high turnover group than it is for the low turnover group. 55 Table 20 –Results associated to the regression models for firms sorted by turnover (see 18 and 19 ) Table 21 –Results associated to the regression models for firms sorted by sector (see 16 and 17) Comments for Results associated to the regression models for firms sorted by sector/turnover. Above it is easily noted that only two stocks demonstrate a negative average return over the complete period, Astra Zeneca and Telia Sonera. Of more interest the model specification using, the 1 month T-bill, the term spread –the difference a 10 year treasury bond and the 3 month T-bill, and the oil price using a volume filter is modest. The explanatory power for the whole model is, not unexpectedly, best for the energy sector with Lunding Petroleum, followed by the materials, industrials and financials respectively. 56 Table 22 - The average value for each explanatory variable in the three volume groups This table provide insights of sign and value of each explanatory input variable. An interesting finding is that on days with medium and low volume the average return in the oil price is positive for 26 and 27 of the firms respectively, while on high volume days it is only positive for 30 % of the firms on average. 57 Table 23 – OMXS price graph with the return, and high and low volume days 1000 OMXS30 Low volume High volume Price 10.0% 6.0% 2.0% 100 -2.0% R -6.0% -10.0% 1/2/1991 1/2/1992 1/2/1993 1/2/1994 1/2/1995 1/2/1996 1/2/1997 1/2/1998 1/2/1999 1/2/2000 1/2/2001 1/2/2002 1/2/2003 A strong tendency though out the 20 year sample periods of volume clustering. 58 1/2/2004 1/2/2005 1/2/2006 1/2/2007 1/2/2008 1/2/2009 1/2/2010 1/2/2011 Table 24 – A graph of the explanatory variables 15 1000 12.5 10 100 7.5 5 1M T-Bill, TB Term Spread, TS Oil Price, OP 2.5 10 0 -2.5 -5 1 The interest rate and oil price movements over the 20 year sample period. 59
© Copyright 2026 Paperzz