Foreign ownership on the Swedish stock market

Transcription

1 DEPARTMENT OF ECONOMICS Uppsala University D-level thesis Author: Petter Holm Supervisor: Martin Holmén Spring 2006 Foreign ownership on the Swedish stock market -what is the attraction of financial ratios on investments from abroad?

2 Abstract Investors in the financial market are supposed to hold diversified portfolios to minimize their risk adjusted for expected return. However, several researchers have pointed out that most investors are over weighted in their home market. This means that most diversification happens in terms of choosing stocks in the home market which means that further possible diversification through international diversification is unused. One can therefore expect that foreign investors have preferences for securities with specific characteristics once they go abroad. An earlier study of the Swedish stock market over the years has shown that foreign investors, in greater extent than domestic investors, have a preference for large firms, firms paying low dividend and firms with low leverage. With the steep up-turn of the Swedish stock market before the millennium and the down-turn in year 2000 in mind, this study examine whether the investment patterns between 1996 and 2005 are consistent with the results of earlier investigations. In general the results are consistent with earlier investigations. However, this study also shows that foreign investors seem to be more interested in choosing securities with relatively high fundamental value and lower level of leverage during market down-turns. Keywords: Sweden; Home bias; International diversification; Foreign investors

3 1 Introduction Theory, earlier studies and empirical implications Theory Earlier empirical studies Empirical implications Data and descriptive statistics Methodology Results Conclusions... 19

4 1 Introduction According to classical portfolio theory, investors should hold a portfolio with minimum risk adjusted for expected return. However, several studies discover a Home Bias-effect, which means that there is a tendency that investors are keener on holding securities within their home market (see for example Lewis (1999) for a survey on the home bias-effect). Reasons for this pattern are that investors in general are better informed about their home market which in extension means that they experience a greater risk in making investments abroad. With respect to this pattern one can expect that foreign investors differ from domestic investors in some respects. Dahlquist & Robertsson (2001) find that investors from abroad prefer securities in large firms that pay low dividends and have low debts levels. Furthermore, foreign investors show a preference for firms with a wide spread of ownership and firms that participate in the international markets to a larger extent. Dahlquist & Robertsson make their research based on ownership in Swedish firms during the period. It is widely known that the Swedish stock market rose dramatically in the last years before the millennium and fell in a steep way thereafter. During the last three years the stock exchange has risen again. This leaves the question if the preferences of foreign investors have changed in the years after This study is supposed to find out whether foreign investors on the Stockholm stock exchange, compared to domestic investors, have different preferences when they pick securities. More precisely the study will deal with the following questions: (i) (ii) What is the relation between a number of financial ratios and the degree of foreign ownership? In what way, if any, have the investment patterns in Swedish stocks made by foreigners changed during different market conditions? The period of research stretches from 1996 to 2005 and includes yearly observations made in the end of each year. The starting year is chosen to generate a practical time length of ten years and to catch considerable amount of the stock market appreciation before the millennium. This study is based on the fifty securities at the Stockholm stock exchange (OMX) that had at least 25 percent of foreign ownership in the end of Some of the 1

5 companies have been purchased or delisted which means that there is a reduction of included observations during the research period. Data will be described more in detail in chapter 3. Dahlquist & Robertsson (2001) argue that the type of investor in the meaning of institutional or individual investors has an implication on investment patterns. Owing to restriction in time and access to data, this study will not be able to separate different types of investors and foreign investors from different geographic regions. In other words, this study will only give a hint of different key ratios implication on the degree of foreign ownership and if the patterns have changed during different market conditions. The structure of this thesis is as follows: Chapter 2 gives an introduction in the theory of how international investors are supposed to pick their stocks. The result of some earlier results is presented followed by hypothesises of mine regarding the implication of firm specific characteristics on foreign ownership. Chapter 3 describes the data used in the study. Chapter 4 goes through the statistical methodology of panel data. Chapter 5 presents and analyses the results of the tests. Chapter 6 summarizes the main conclusions of this study. 2

6 2 Theory, earlier studies and empirical implications 2.1 Theory The classical portfolio theory says that investors are risk averse and choose a portfolio with maximum risk adjusted return. The relationship between risk and expected return for a security is given by Capital Asset Pricing Model (CAPM), which is a well known model in the field of financial theory (Bodie et al. 2002). If all investors are assumed willing to optimize their expected return adjusted for risk, every efficient portfolio will lie on a straight line called Capital Market Line (CML) which shows the trade off between expected return and risk. Minimizing the risk of the portfolio given the expected return is called diversification. Diversification is possible when the covariance between securities in the portfolio is low or even negative. Expressed in more familiar terms one can say that the variation of the individual securities tend to follow different patterns and in extension this means that the risk of the whole portfolio becomes lower. The assumption that investors pick securities according to the trade-off between risk and expected return and how the individual security relates to the rest of the stocks in their portfolio is not the case for most investors in reality. Merton (1987) among others argues that with imperfect information the classical theory is not necessary valid when investors have different information. Merton explain in a theoretical way that institutional structures and cost to get information lead to considerable number of investors acting that traditionally is seen as divergence from an optimal portfolio assumption. Further on Merton stipulates that investors are familiar with different subsets of stocks and consequently use stocks of their knowledge when they construct their portfolios. Merton motivates this with the fact that both institutional and individual investors hold only a small fraction of the securities available even if he admit that other factors like transaction costs, taxes and limited divisibility have influence. The knowledge of a security is supposed to be related to in what extension professional analysts are following the security. Ross et al. (2002) present some topics of statement analysis for individual shares. One aspect of firms is the ability to meet short term payments. One measure for this is the current ratio which is the ratio between current assets and current liabilities. Ross et al. point out that the 3

7 current ratio should be measured over several years and compared with other firms that are in similar industries. Another aspect is to which extent a firm rests on debt financing. One measure of the degree of leverage is the ratio between total debt and total equity. The degree of debt can be viewed from two opposite directions. One positive effect of more debt is that debt financing is tax deductible. In addition there is a possible leverage effect if further debt financing generates a return in the business of the firm compared to the borrowing interest rate. However, more debt financing can create a negative effect called financial distress, which means that it can reduce the ability to manage demand for payment and in extension cause problem to obtain additional loan for financing. Book-to-market ratio is the ratio between market price of all issued shares and the equity of the firm (book value). Low bookto-market value is often seen as a safer investment, depending on the fact that the book value is seen as a floor for the market price. The reason is that the asset of the firm can be sold. This is however a topic of discussion, but there is a general opinion that low book-to-market ratio is associated with some safety (Bodie et al. 2002). 2.2 Earlier empirical studies Several studies have been made in the field of international diversification and home bias assumptions. Tesar & Warner (1995) find that investors are keener on taking positions in their home market and therefore they do not fully utilize the potential gains with international diversification. Tesar & Warner make their study with focus on the investment patterns in Canada, Germany, Japan, UK and the USA during the period One main conclusion in their study is that investors have increased their international positions during the period of research but there were still unutilized potential gains from international diversifications in the end of the period. In addition Tesar & Warner finds that foreign investors hold securities with higher liquidity, which is measured as the turnover rate. Kang & Stulz (1997) study the patterns of foreign ownership on the Japanese stock market between 1975 and Among other things they find that foreign investors, compared to domestic investors, are keener on holding securities in firms with low debt, bigger turnover of the shares, and higher export rate. These are results that are consistent with the study of Dahlquist & Robertsson (2001). French & Poterba (1991) investigate the preferences of investors regarding holding securities in foreign markets. A general result is that investors hold disproportionately more shares in 4

8 their home market. French & Poterba make the conclusion that the main reason for this pattern is investor preferences and not institutional regulations or transaction costs. Their opinion of transaction cost as insufficient in explaining home bias pattern is confirmed by Tesar & Warner (1995). The fact that investors know less about firms and market conditions abroad may imply that they value risk associated which such investments higher. Further on, Shiller et al. (1990) compare forecasts done by professional investors in United States and Japan who predicts the future development of the stock market in the two countries. Investors in both countries tend to forecast a relative stronger future development of the stock market in their home country compared to the foreign country. Strong & Xu (2003) investigates the investment pattern of fund managers and find that they are heavily over-weighted in their home market compared to the home market share of total world market. Home bias pattern is therefore the case even for professional fund managers. Dahlquist & Robertsson (2001) find that foreign investors prefer securities in large firms that pay low dividends and have low debts levels. Furthermore, firms with a wide spread ownership tend to attract investments from abroad. Another factor affecting foreign investors is the presence in the international markets which Dahlquist & Robertsson measure through the export sales and if the firm is listed on a foreign stock exchange. The result is that that foreign investors prefer to invest in firms with high export rate and firms that are listed on a stock exchange outside Sweden. 2.3 Empirical implications From the theoretical issues and earlier empirical result that have been illustrated above my research hypotheses will finally be presented. The hypothesises will in general rest on the assumption that foreign investors are expected to be less informed about the Swedish stock market compared to domestic investors. One implication of the theory by Merton (1987) is that investors choose securities in bigger firms when they go abroad since bigger firms are more covered by media and financial analysts. Thus, bigger firms are better known in the international market. This hypothesis is also mentioned by Dahlquist & Robertsson. In addition to firm size, listing on foreign stock exchange will be used as a proxy for foreign investor s knowledge of the firm. The fact that some foreign countries including the United States have a heavier tax burden on dividends compared to the tax on capital gains, leads to the expectation that foreign investors will prefer to invest in firms with restrictive dividend policies that instead generate better returns for their shareholders. Litzenberger & 5

9 Ramaswamy (1979) find that investors in USA demand higher before tax return if the increasing return is in the form of dividends. Implications of the financing structure of firms were presented in the theory section. There is however hard to have a clear opinion about the implication of debt financing on foreign ownership. The research hypothesis in this study will rest on the assumption that foreign investors are less informed and therefore have a preference for firms with relatively lower debts which they evaluate as safer holdings. The information assumption will also serve as a base for the hypothesis in this study stipulating that investors from abroad expects to have a preference for stocks with lower beta (systematic risk). Foreign investors are also supposed to have a preference for firms with lower book-to-market ratio, which they judge as a safer investment when fundamental values are relatively higher. With the study of Tesar & Warner in mind the hypothesis in this study is that turnover rate is positively related to foreign ownership. The share of total votes by the largest owner is used to see if foreign owners are interested in management of the firm. The expected hypothesis is taken directly from Dahlquist & Robertsson (2001) and the assumption used is that foreign owners interested in management are expected to avoid firms with high owner concentration. Before going on with the study this section presents a summary of the hypothesis stated above. Foreign investors are supposed to be relatively more attracted to: (i) bigger firms and firms listed on a foreign stock exchange, (ii) firms with restrictive dividend policy, (iii) firms with lower debts, (iv) firms with lower beta, (v) firms with lower book-to-market and (iv) firms with higher turnover. No hypothesis was presented for votes by largest owner. The assumption is that foreign investors avoid firms with high concentration of the ownership if they are interested in management. 3 Data and descriptive statistics This study includes all fifty companies listed on the Stockholm stock exchange that had at least 25 percent of foreign ownership in the end of It is therefore not the fifty firms with the biggest foreign ownership counted in capital value. This selection is stationary 1 The original set of firms is picked from Sundin A. & Sundqvist S-I., Observations from the following years research are based on data from the books in the same series that are issued every year. 6

10 through the ten year period which means that the original fifty stocks serves as base for the study even if other companies reaches 25 percent of foreign ownership during the period. Included companies being bought or delisted creates some reduction of observations in the data. As can be seen in table 1 (page 7) the effect of the reductions is that only 19 out of 50 original companies is still noted on the Stockholm Stock Exchange in the year In total the decline in data of foreign ownership are 196 observations due to delistings, which leaves 304 observations serving as a base for this study. However, missing observations in the dataseries of firm specific characteristics provided by Datastream imply that observations are actually used in the panel regressions. Table 1: Foreign ownership in percent of the firms total market capital* Firm ABB 35,6 31,7 42,8 Acrimo 57,2 Aga 38,6 36,4 46,6 Allgon 34 23,6 12,4 20,3 28,2 16,6 18,4 Argonaut/N&T Argonaut 27, ,6 ASG 36,4 10,6 26,3 Assa Abloy 74,7 67,7 60, ,1 55,9 57,4 48,8 35,3 27,8 Astra 44,5 40,4 39,4 Atlas Copco 38, , ,9 34,3 41,7 45, ,3 Autoliv 59,4 Avesta Sheffield 66,4 66,1 66,2 72,2 BT Industries 26,9 24,2 16,5 15,7 BTL 50,8 53,7 85,6 Bure/Bure Equity 39, ,6 20,6 33,9 13,7 14, ,7 11 CynCrona 48,2 Dahl 53,8 53,9 45,6 Electrolux 50, ,9 49,5 41,7 41, ,8 37,9 43,7 Ericsson 51, ,7 55,4 54, ,5 43,9 45,9 45,8 Esselte 42,8 35,6 28,5 25,2 15,7 20,4 Fjällräven/Fenix outdoor 68,3 71, ,2 79,4 72,7 72,5 69,8 69,6 87,7 Frontec/Acando Frontec 45,1 19,5 19,9 16,1 12,2 14,3 11, ,2 14 Frontline 82,8 Getinge Industrier/Getinge 38,1 25,1 22,4 14,5 15,8 32,7 33,2 34,9 35,3 28,8 Graninge 49,4 51,3 51, ,7 55,9 55,8 Gullspång 44,5 Höganäs 49,1 37,6 24,5 20,4 32,2 32,1 25, ,7 22,2 ICB Shipping 63,6 73,9 77,7 Kalmar Industries 48,2 72,6 69,2 70,9 Kinnevik 36,4 24,1 24,1 16,6 17,2 11,3 10,5 11,6 12,9 11 Lindab 73,3 73,9 63,8 57,3 52,3 Lindex 27,6 27,9 21,2 22,5 23,8 24,3 19,7 21,9 44,8 59,4 Munksjö 68,3 62, ,9 55,6 51,1 Nobel Biocare 67,2 56,3 56,4 59,9 90,4 88,4 Nordic Tel/Europolitan 79,4 80,1 82,9 81,8 78,4 75,9 80,7 Nordström&Thulin 43,8 OM Gruppen/OM/OM Hex/OMX 30,5 33,1 27,7 28,5 24,2 18,2 10,8 24,1 12,5 18,5 Pricer 42,6 51,1 27,9 1,8 22,1 23,9 22,1 20,3 22,7 25,1 Sands Petroleum/Lundin oil 48,5 46,4 37,3 37,8 47,4 Scanem 38,3 36,4 33,9 Securitas 51,2 46,9 52,7 59, ,9 56, ,3 43,7 SinterCast 87,6 51,9 46, ,5 37,6 44,7 46,1 44,3 39,4 Skandia 62,2 70,4 61,9 61,4 55,4 40, ,2 46,4 72 SKF 39,7 39,5 36,6 44,2 36,4 28,8 39,4 51, ,8 Solitair 77,5 71 Stora 31,7 26,4 Svedala ,7 Swedish Match 37,5 47,8 52,6 49,6 52,4 68,5 63,8 60,2 72,6 75,9 Sydkraft 35,8 54, ,3 78,2 The Empire 27,4 25,3 25,4 23,5 25,2 25,1 25,5 Volvo 42 36,4 37,9 28,1 39,6 44,1 47,8 51,8 51,4 49 * The table shows the foreign ownership in percent of total market capital in each firm between 1996 and More than one name is stated in some cases where the name of the firm has changed during the period. It is obvious that some firms only have a limited number of observations. The reason is that these firms have been bought by other interests or delisted from the stock exchange. In total there are 196 missing observations and therefore 304 observations are left and serve as a base for this study. 7

11 As mentioned in the introduction, market conditions of the Swedish stock market have changed considerable over the ten year period which this study examines. Figure 1 (page 8) makes a comparison between the development of Affärsvärldens generalindex 2 (left figure) and the percentage ownership of total stock exchange value hold by foreign investors (right figure). There seems to be some connection between the development of the Swedish stock market and investments from abroad even if it is difficult to determine which factor affects the other. The large share of foreign ownership on the Stockholm Stock Exchange has also the implication that it becomes more connected to the movements of international stock exchanges. Foreign investors are assumed to reduce their ownership on distant markets in times of disturbances in their domestic market. 3 However, the main topic for this study is that it gives a guidance of suitable dividing in shorter periods when to investigate whether investing patterns tend to change during different market conditions. 450 Affärsvärldens generalindex Foreign ownership in percent Figure 1: The left figure shows the index development for Affärsvärldens generalindex over the period measured on daily basis each stock exchange trading day. Right figure shows the percentage ownership of total stock exchange value hold by foreign investors measured in the end of each year. Foreign ownership in the firms in this study will be tested with respect to a number of financial attributes of the firms. Mainly these attributes are the same as in the earlier study by Dahlqvist & Robertsson (2001). However two variables are excluded. Residual variance was insignificant in their study and is therefore excluded. As a consequence of insufficient access to data, the export rate variable is also excluded. Included variables in the study are presented below: 2 Affärsvärldens Generalindex is a financial index that is commonly used in Sweden as benchmark and a value weighted index that measures the general course development on the Stockholm Stock Exchange. 8

12 (i) Market value: Market value is measured by the total value of all issued shares in the firm. As in some earlier studies in this field the log of the market value will be used in regressions (se for example Kang & Stulz 1997 or Dahlquist & Göransson 2001). (ii) Dividend yield: Dividend yield in percent is measured as the value of all dividend paid during the year divided by the market value of the company at the beginning of the year. (iii) Return: Return on shares is generated from a market price index on weekly basis. Yearly return is thereafter calculated by taking the return over a period of 52 weeks. (iv) Systematic risk: Systematic risk is given by historical betas in the program Datastream. Benchmark is therefore not the same as in the study by Dahlquist & Robertsson (2001). However one can expect that the relative beta level between individual firms is of more importance than the actual level. (v) Book-to-market: Book-to-market is generated by multiplying book value per share and number of issued shares and then dividing it with equity of the firm. (vi) Current ratio: Current ratio shows the ratio between current assets and current liabilities. Current ratio has been generated direct from Datastream (vii) Leverage ratio: Leverage ratio is also named debt-to-equity ratio and is the ratio between total debt of the firm and total equity. (viii) Return on equity: Return on equity is measured by net income divided by total equity. Return on equity is generated directly from Datastream. (ix) Turnover rate: Dahlqvist & Robertsson use turnover rate as a measure of liquidity of the firm s shares. It is measured as total value of traded stocks in one year divided by the market value of the firm. 4 (x) Concentration: Share of total votes in percent hold by the largest shareholder. (xi) Foreign listing: A dummy variable with value 1 if the stock is listed on a stock exchange outside Sweden and 0 if otherwise. 3 Article Utländska aktieägare gör börsen känslig in the Swedish newspaper Dagens Nyheter Data of total value of traded stocks is picked from OMX group, Stockholm Stock Exchange Internet page. OMX group only provides historical data of trading volumes from year 2000 on their Internet page, which means that the turnover variable suffers from considerable amount of missing data. 9

13 Table 2 (page 10) shows descriptive statistics for the variables used in this study. A brief overview indicates that there is considerable variation in all parameters which one can expect when the sample consists of firms from different business sectors and sizes. Mean value for market value is about five times higher than the median value. This means that there are a few observations with high market values that push the mean value upwards, while most firms in the sample have lower market values. Implications are similar for other parameters with mean value greater than median value and vice versa if the median value is higher. In general, descriptive statistics seems reasonable. However, return on equity has a lowest observation of more than minus one hundred percent which is surprising. Verifying with the original generated return on equity series confirm that there is two observations of a negative return on equity of more than one hundred percent. This implies the possibility that the surprising observations are generated when owners have raised more equity and done further losses during the year. The original observations are therefore used without adjustments. Table 2: Descriptive statistics for included variables* Variable Mean Median Highest Lowest Std. Dev. Observations Foreign ownership 43,41 43,35 90,40 1,80 18, Market value 39414, , ,00 88, , Dividend yield 2,33 2,03 59,70 0,00 3, Return 17,69 11,13 278,34-96,63 53, Systematic risk 0,87 0,77 2,72-0,52 0, Book-to-market 0,43 0,32 2,98 0,00 0, Current ratio 2,66 1,69 71,00 0,51 5, Leverage ratio 0,61 0,53 2,57 0,00 0, Return on equity 10,13 15,07 215,56-107,78 29, Turnover rate 1,15 0,94 7,73 0,01 1, Concentration 28,68 25,10 85,50 4,40 17, Observations with foreign listing Observations without foreign listing Foreign listing * Foreign ownership is given in percent of the firms total market capitalization. Market value is stated in million SEK. Dividend yield and return and return on equity are in percent. Systematic value is the yearly beta-value for each stock. Book-to-market is the total book value for all shares divided by market capitalization. Current ratio is current assets divided by current liabilities. Leverage ratio is total debt divided by total equity. Turnover rate is total value of traded securities traded over a year divided by the market value of the firm. Concentration is the percentage share of total votes holding the actor with biggest share of total votes in each firm. Foreign listing shows how many of the observations that are listed on a stock exchange outside Sweden and take dummy value 1. Observations without foreign listing correspond to dummy value 0. The study includes yearly observations between 1996 and 2005 for all time series except turnover rate which stretches between 2000 and Descriptive statistics is generated on the number of observations with observed values for all firm specific characteristics. The number of observations is somewhat higher in some regressions when variables with more missing variables are omitted. 10

14 The panel data consist of both a time and a cross-sectional dimension. A general topic is then to test if the individual time series are stationary. Wooldridge (2003) emphasize that nonstationary series can cause spurious regressions. The big extent of time series in this study combined with small number of observation for each series has lead to the decision to ignore this possible problem. The fact that many variables are ratios and shares along with capitalization value that follows the market situation can eventually reduce this problem. Dahlquist & Robertsson (2001) do not test if series are stationary which support the decision. 4 Methodology Panel data can be used when to study the same sample of individuals over a period of time (Pindyck & Rubinfeld 1998). One main advantage with panel data is that the test will consist of more degrees of freedom. In everyday expression it can be explained as that the number of data observation points increases. A panel data regression is a combination of cross-sectional data and time-series data. Two versions of panel data are the cases of fixed-effect model and random-effect model. The Fixed-effect model uses dummy variables to take changing intercepts in cross-series and time-series into consideration. The fixed-effect model can be written as follows (see equation 1): Y it = α + βx + γ L + it 2 W2t + γ 3W3 t + L + γ NWN + δ 2Zi2 + δ 3Zi3 + + δt ZiT ε it (1) where α is the intercept, β is the slope coefficient, W is dummy variable for individual i, Z is a dummy variable for time period t and ε is the error term for individual i in time period t (Pindyck & Rubinfeld 1998). The Random-effect model does not use any random variables in order to vary the intercept. Instead the intercept become a random variable and takes a mean value. The Random-effect model can be written as expressed in equation 2: Y = α + β (2) it X it where parameters should be interpreted in the same way as in equation 1. Baltagi (2001) says that a fixed-effect model should be used when the sample is a specific set of firms which is the case in this study. However, there is a statistical test called Hausman-test 11

15 that allows testing whether random-effect is an appropriate model to use or if the data of the study requires a fixed-effect model. If the sum squared residuals differ substantially between the two models, fixed-effect model should be used. The Hausman test statistic is written as follows (see equation 3): ( SSRur SSRr ) ( N + T 2) FN + T 2, NT N T = (3) ( SSRr ) ( NT N T ) where SSR r and SSR ur are the sum square residuals for fixed-effect and random-effect models, N is the number of cross-sections and T is the number of time periods. The Null hypothesis (H 0 ) is that the random-sample is a suitable model to use. Otherwise, the fixedeffect model with varying intercept should be used (Pindyck & Rubinfeld 1998). Readers with a deeper interest in statistics and econometrics are recommended to read about panel data in for instance Arellano (2003), Baltagi (2001) or Pindyck & Rubinfeld (1998). 5 Results The objective of this study is to give an answer to the following questions: (i) What is the relation between a number of financial ratios and the degree of foreign ownership? (ii) In what way, if any, have the investment patterns in Swedish stocks made by foreigners changed during different market conditions? To answer these questions panel regressions are estimated between the foreign ownership and some basic key ratios over the year To decide if there is some information in the selection of firms, ordinary least square (OLS) regressions are estimated and compared with panel regressions estimated with fixed-effects. To see if panel data estimation with random-effect is sufficient for the sample or if fixed-effect is to prefer a Hausman-test is done over the period To see if the investment patterns have changed during different market conditions the sample period is divided into three shorter periods. With the development of Affärsvärldens generalindex and foreign investors ownership that were illustrated in figure 1 in mind the periods are chosen to be year , and Since knowledge of the firm, liquidity of the shares and voting power diversification are expected to be factors having implication on foreign investors, variables counting for these effects are then used as regressors. 12

16 Table 3 (page 15) shows the result of panel regressions estimated by fixed-effects and regressions estimated with OLS. For many variables there is a difference regarding that they are significant with one of the regression methods. Focusing on variables that seem to have significant impact on foreign ownership with both methods, one can see that Logarithmic market capitalization shows a positive relationship for the year This result is consistent with the hypothesis presented in chapter 2. Estimates for market capitalization can be viewed as that the size of the firm has larger implication on foreign investments for the period if judging from estimation with fixed-effect. However it is hard to decide if the result depends on changing relationship or the decreasing sample between the periods. Especially when there is a considerable difference between the two up-turn periods of the Swedish stock market as well as between the results generated by fixed-effect and OLS. Dividend yield is negative related to foreign ownership with five percent significance for year with both estimation methods and beta shows a significant negative relation with foreign ownership over periods and Foreign investors seem to hold a larger share in securities with lower systematic risk (beta). Estimation for leverage ratio confirms between the estimation methods for period and shows that foreign owners seems to be relatively more attracted to firms with lower leverage. Mentioned results for dividend yield, beta and leverage ratio corresponds to the expected relationships with foreign ownership stipulated in chapter 2. Judging from the Wald tests, the parameters in the fixed-effect model together seems to have a statistically greater impact in explaining the foreign ownership. For some variables the coefficients generated with OLS are considerably larger than with fixed-effects. This can be due to the fact that OLS credit these variables a greater importance, while it actually can be a result of the selection of companies and that these companies have particular qualities. Focusing on the fixed-effect estimations one can see that additional variables are significant for some periods. Return and book-to-market is significant for some period. The negative coefficient for book-to-market over the period corresponds to the hypothesis that was presented earlier. The fact that this result is during the down-turn of the stock market after the millennium can possibly be explained by a more defensive investment strategy taken by foreign investors. It can be interpreted as that foreign investors in the period invested in firms with share prices that were built upon future expected returns and in the downturn changed to companies with higher fundamental value compared to 13

17 aggregate share price. Current ratio is the only parameter that is significantly different from zero for all periods. The highest positive estimate is for the period A guess could be that investors became more concerned about fundamentals and ability to manage short term payments in the downturn of the Swedish stock market during these years. For shorter periods, no relation can be seen between return on equity and foreign ownership. But over the long period there is a negative relationship between the two variables. However, it can be noted that even if the coefficient is statistically different from zero, it has only a small impact on the foreign ownership. Wald test p-values 0.30 for period implies that null hypothesis that the sum of all coefficients are equal to zero can not be rejected. This means that there is a possibility that the model does not provide any information in explaining the degree of foreign ownership. However, judging from low Wald test p-values for other periods and high standard deviation caused by low number of observations for period one can expect that the model in general can explain some variation in foreign ownership. When interpreting the results of the panel regressions it should be mentioned that the relationship between firm-specific characteristics and foreign investments might work in the opposite direction. It could be the case that foreign investments affect some of the variables in the study and not the other way around. This is an important topic paying attention to in all panel regressions estimated in this study. However, even if the reader should keep the possibility of foreign investors affecting the research variables in mind the discussion will focus on the viewpoint that the firm-specific characteristics might have some implication on foreign investments. To test whether a random-effect model is to prefer compared to the fixed-effect model a Hausman-test is performed over the period The result is that the null hypothesis of random-effect model being appropriate is rejected (se lower part of table 3). This result and the fact that Baltagi (2001) stipulates that fixed-effect models are appropriate to use when the sample is a specific set of firms implies that regressions will be estimated with fixed-effects in the rest of this study. 14

18 Table 3: Regression of foreign ownership on firm-specific characteristics* Fixed OLS Fixed OLS Fixed OLS Fixed OLS Market capitalization 18,14 2,84 26,79-1,31 0,00 8,27 0,00 6,02 0,00 0,06 0,00 0,56 0,76 0,01 0,60 0,18 Divident yield -0,17-0,19-0,25-0,19-2,11-5,10-6,90-0,61 0,27 0,25 0,44 0,19 0,00 0,00 0,03 0,84 Return 1,32 1,62-1,60 4,77-1,99 1,05-1,56 1,96 0,27 0,46 0,00 0,21 0,30 0,90 0,53 0,51 Beta -2,70-12,77-5,71-10,37-8,64-16,94-4,86-11,43 0,08 0,00 0,03 0,01 0,00 0,00 0,60 0,03 Book-to-market 4,20 1,19 7,71 6,91-10,76 4,05 0,52-9,15 0,23 0,72 0,15 0,12 0,03 0,44 0,85 0,09 Current ratio 0,70-0,09 0,48-0,02 2,86 3,89 0,85 0,96 0,00 0,45 0,00 0,89 0,00 0,12 0,03 0,68 Leverage ratio -5,36-4,33-3,22-2,58-6,15-10,58 2,22 1,55 0,00 0,10 0,01 0,44 0,02 0,01 0,55 0,81 Return on equity -0,07 0,05-0,02 0,01 0,11 0,21-0,03 0,03 0,01 0,24 0,41 0,90 0,08 0,02 0,57 0,72 Adjusted R 2 0,80 0,15 0,80 0,05 0,96 0,42 0,88 0,39 Wald test p-value 0,04 0,03 0,03 0,75 0,00 0,21 0,30 0,39 Number of observations Hausman test SSR random SSR fixed F-value 20261, ,03 3,42 Region of rejection F>1,33 * The table shows the estimated coefficients from regression against foreign ownership. Left column under each period shows the estimations generated by fixed-effect panel data while the right column shows the result generated by a common Ordinary Least Squares (OLS) regression. As expressed in chapter 3 the log of market capitalization is used as regressor. First three periods is estimation result for short periods that is chosen to reflect changing market conditions on the stock exchange between these years presents the results from regression over the whole period of research. All regressions are estimated with heteroskedasticityconsistent standard errors. P-values are reported under each coefficient. Wald test p-value is related the null hypothesis that the sum of coefficients are equal to zero. The bottom of the table shows the result from Hausman test. SSR random and SSR fixed are the sum squared residuals from estimations with cross-section random effects and fixed effects for cross-section and time series. The F-value is calculated by substituting the values for SSR random and SSR fixed into equation 3. Region of rejection shows lower limit where H 0 is rejected on 5 percent level of significance. As mentioned in section 2.1 and 2.3 foreign investors are supposed to have a preference for firms that are well known to the investors. Table 4 (page 17) shows the result when using foreign listing as a variable that is expected to represent this preference. Turnover rate and concentration of voting power associated to the biggest owner are used to count for effects of the security s liquidity and the risk with one dominating owner, respectively. Every regression estimated with fixed-effects generates a Wald test p-value less than Market capitalization seems to have a big impact on foreign ownership and dividend yield are negative related as 15

19 expected for some regressions. Book-to-market and current ratio seems to relieve each other in the meaning of generating coefficients different from zero. Estimated coefficients for book-tomarket are however the opposite compared to the expected relationship presented in section 2.3. Coefficients for current ratio on the other hand support the hypothesis of foreign investors being relatively more attracted to firms with better ability to manage short term payments. The estimation results do not support some relation between the liquidity of the shares and foreign ownership. Once again it should be emphasized that turnover-data consist of a limited number of observations. Concentration seems to have a clear impact on foreign ownership if judging from regression (iii) and (vii) which consist of observations from the full period in contrast to regression (v) and (viii) that are base on a shorter time period. The conclusion is however that the relationship seems to be positive between owner concentration and foreign ownership. Dahlquist & Robertsson (2001) present a theory that foreign investors interested in management would avoid firms with high concentration. This result can therefore be an indication that foreign investors are not interested in management or a sign of that biggest owner in some firms in fact are foreign owners. However, it should be pointed out that the shares in many firms listed on the Swedish exchange are classified in different categories according to the voting power of the share. Normally there is a distinction between A- and B- securities where A-securities in general are issued in less extent. This means that one big owner holding A-securities may have a share of total votes that is considerably higher than the share of the equity. None of the foreign listing estimates are significant at five percent level. If the significance level are stretched to ten percent two coefficients becomes statistically different from zero. Surprisingly these coefficients are negative implying that foreign listing is negatively related to foreign ownership which is the opposite of the stated hypothesis. However, a possible explanation is that the relationship may work in the opposite direction where firms with less foreign investments list their share on a foreign exchange to attract investors. 16

20 Table 4: Regressions with variables serving as a proxy for investors knowledge of the firm* (i) (ii) (iii) (iv) (v) (vi) (vii) (viii) Market capitalization 19,27 28,91 19,70 19,70 28,87 29,12 20,09 29,18 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 Dividend yield -0,08-1,96-0,19-0,12-1,95-2,16-0,24-2,17 0,52 0,02 0,13 0,28 0,02 0,02 0,05 0,02 Book-to-market ratio 2,71 7,18 3,22 3,29 7,18 7,72 3,75 7,73 0,40 0,03 0,32 0,30 0,03 0,01 0,24 0,01 Current ratio 0,69 0,10 0,91 0,68 0,10-0,32 0,90-0,32 0,00 0,88 0,00 0,00 0,88 0,63 0,00 0,63 Turnover rate -0,41-0,42-0,35-0,34 0,53 0,54 0,53 0,54 Concentration 0,41 0,02 0,41-0,02 0,00 0,88 0,00 0,80 Foreign listing -2,73-6,72-2,51-6,76 0,07 0,19 0,10 0,19 Adjusted R 2 0,79 0,88 0,81 0,79 0,88 0,89 0,81 0,89 Wald test p-values 0,00 0,00 0,00 0,00 0,00 0,00 0,00 0,00 Number of observations The table shows the estimation results of regressions between foreign ownership and different firm characteristics. Regression (i), (iii), (iv) and (vii) are estimated over the years , while (ii), (v), (vi) and (viii) have a sample period between 2001 and 2005 owing to limited access to turnover data. Heteroskedasticityconsistent standard errors are used in all regressions and p-values are reported under each coefficient. Wald test p-value is related to the null hypothesis that the sum of explaining variables coefficients is equal to zero. To see if the if the relationship result between proxy variables used in table 4 and foreign ownership is stable, new regressions are estimated by taking significant variables over the period in table 3 and combine them with the same proxy variables as done in table 4. Beta has a p-value of 0.08 but is included in these regressions even if 0.05 often is used as significance level. The regression results in table 5 (page 18) show that market capitalization has a significant coefficient in all regressions which is consistent with the stated hypothesis. Beta has a significant negative coefficient in regression (i) and insignificant negative estimates in regression (ii) (viii). This can indicate that foreign investors have some tendency of choosing stocks with lower beta value as was expected according to the hypothesis. Further on, a higher degree of foreign ownership seems to be related to lower leverage, judging from the coefficients of current ratio and leverage ratio. The hypothesis is that lower leverage is associated with lower risk and therefore attracting less informed foreign investors. Investors from abroad tend to have lower return on their equity according to the estimation result. No exact hypothesis was stated for the return on equity, but these results indicate that foreign investors seem to invest in lower return on equity. The only proxy variable with any estimates significant different from zero is owner concentration. As in the regressions 17

21 presented in table 4 the relationship is between concentration and foreign ownership is estimated to be positive. As mentioned before, this implies that foreign investors are not interested in management of the firm. Table 5: Regressions with variables serving as a proxy for investors knowledge of the firm* (i) (ii) (iii) (iv) (v) (vi) (vii) (viii) Market capitalization 15,24 14,05 14,93 15,29 14,04 14,41 14,97 14,40 0,00 0,04 0,00 0,00 0,04 0,04 0,00 0,04 Systematic risk (Beta) -3,87-3,19-3,04-3,59-2,62-2,30-2,87-1,97 0,05 0,57 0,10 0,10 0,73 0,69 0,17 0,80 Current ratio 0,78 0,84 0,99 0,78 0,80 0,65 0,99 0,63 0,00 0,06 0,00 0,00 0,12 0,27 0,00 0,34 Leverage ratio -4,34-5,74-4,73-4,29-5,80-5,14-4,70-5,18 0,02 0,05 0,01 0,02 0,04 0,06 0,01 0,05 Return on equity -0,06-0,05-0,05-0,06-0,05-0,06-0,05-0,05 0,02 0,09 0,02 0,02 0,16 0,07 0,02 0,12 Turnover rate 0,03 0,02 0,07 0,06 0,96 0,98 0,92 0,93 Concentration 0,37 0,05 0,37 0,03 0,00 0,82 0,00 0,88 Foreign listing -0,96-3,50-0,60-3,46 0,60 0,34 0,74 0,33 Adjusted R 2 0,79 0,87 0,81 0,79 0,87 0,87 0,81 0,87 Wald test p-value 0,08 0,61 0,05 0,10 0,63 0,53 0,05 0,56 Number of observations * The table shows the estimation results of regressions between foreign ownership and different firm characteristics. Regression (i), (iii), (iv) and (vii) are estimated over the years , while (ii), (v), (vi) and (viii) have a sample period between 2001 and 2005 owing to limited access to turnover data. Heteroskedasticityconsistent standard errors are used in all regressions and p-values are reported under each coefficient. Wald test p-value is related to the null hypothesis that the sum of explaining variables coefficients is equal to zero. 18

22 6 Conclusions The purpose of this study was to examine the relation between a number of financial ratios and the degree of foreign ownership. Beyond that another aim was to investigate if the relations between the degree of foreign ownership and firm-specific characteristics have changed during different market conditions in the Swedish stock market. Foreign investors show a certain preference for securities in firms with high market value. The result is not obvious for all time periods but a general pattern. There are also some indications of a negative relationship between the dividend yield and foreign ownership. These results confirm the hypothesis and results in earlier studies. Investors from outside Sweden also seem more attracted to firms with a lower systematic risk, which one can expect with their restricted ability to follow the firms and get information compared to Swedish investors. There are also indications that foreign investors take larger shares in firms with smaller debts compared to their equity and in firms that seems to have better ability to manage current payments. In general these results support the findings by Dahlquist & Robertsson (2001). Dividing the whole period of research into smaller time periods shows a variety in the estimation results between the time periods. Estimations for book-to-market and current ratio indicate that foreign investors seem to be keener on investing in firms with relatively high fundamental value and low leverage. Reasons behind this tendency can be that foreign investors experience lower risk when holding securities in such firms and that this factor becomes more important during market down-turns. It is however hard to make a general conclusion that can be associated with changing conditions in the financial market. Further adding of variables that are supposed to catch effects of liquidity of the security, international presence and the impact of one strong owner generates some opposite results compared to Dahlquist & Robertsson (2001). A possible reason is that this investigation rests on another selection of companies and therefore it can be the case that the fifty firms in this study with included dropouts show a different pattern than the whole population of firms. The fact that the selection is based on the firms having largest share of foreign ownership in 1996 can also affect the result. Some years thereafter the IT boom reached its peak and that could possibly have affected the sample to consist of relatively less processing and manufacturing industry. 19

23 Processing and manufacturing industry normally consist of higher fundamental value and the regressions for shorter periods indicated a change in foreign investor preferences during year toward firms with higher fundamental value. This and other topics can result in that the selection will consist of firms with different financial characteristics depending on the starting year. The fact that the sample is picked by choosing only the firms with at least 25 percent of foreign ownership can also create a selection bias. A presumably more statistical correct way would have been to randomly draw fifty firms or even better include the whole population of firms. The first obvious topic for further research is to make a total investigation similar to the earlier study by Dahlqvist & Robertsson. Along this study it has been more apparent that the decreasing number of observations can be a problem when drawing conclusions of foreign investors preferences. With the limited access to data through the sources used in this study in mind, a future study can therefore try to receive data from the same data providers that has been used in the earlier study. An alternative way could have been to pick the 22 companies in this study that are listed on the Swedish stock exchange over the whole ten year period. This will however cause a considerable reduction in the number of observations, which basically was the reason to use the original fifty firms. Dahlquist & Robertsson propose that a future subject can be to evaluate how foreign investors succeed in their investments compared to domestic investors on the Swedish stock market. A more difficult topic would be to evaluate the costs of asymmetric information between foreign and domestic investors. 20