Asset Pricing and the Inflation Exposure Risk Factor using Individual Stock Returns and Industry Sector Portfolio Returns: Evidence from the Swedish and Dutch Capital Markets

(1)

Master Thesis

Double-degree MSc BA: International Financial

Management

Topic:

Asset Pricing and the Inflation Exposure Risk Factor

using Individual Stock Returns and Industry Sector

Portfolio Returns: Evidence from the Swedish and

Dutch Capital Markets

University of Groningen

Faculty of Economics and Business (FEB) Uppsala University

Företagsekonomiska institutionen (FEK)

(2)

Keywords: Asset Pricing, CAPM, FF3FM, Market Anomalies, Inflation Exposure, Risk, Cost of Equity

ACKNOWLEDGEMENTS

It is a great pleasure to thank everyone who helped me write my thesis successfully. In particular, I would like to thank my supervisor, Auke Plantinga, for guidance and constructive advice. In addition, I am truly indebted and thankful for all the support that I have received from my family as well as friends throughout my studies.

Abstract

Using individual stock return data as well as industry sector return data from the Swedish and Dutch capital markets over the period January 2006 to September 2012, this thesis project investigates the importance of a novel inflation exposure risk factor to explain asset return variations. Accordingly, a zero-investment portfolio (HILO) is constructed to mimic the risk related to an asset’s inflation exposure. Whereas supporters of the CAPM claim that all risk of an asset can be expressed through the market-‘Beta’, the novel HILO risk factor is tested as an extension of the original Fama and French three-factor model, which already takes into account two other widely observed market anomalies: the size effect and value effect. The results of the conducted regression analysis and the redundant variables test provide support for the inclusion of the HILO factor in the model for both capital markets. However, the evidence is more convincing for industry sector returns than individual stock returns. Next to contributing to a recent stream of academic asset pricing literature, these results assist financial managers and investors in a more accurate estimation of a company’s cost of equity or industry’s cost of equity. In addition, the findings suggest that investors should adjust their investment strategies and hedging strategies according to the desired level of inflation exposure risk.

(3)

1. INTRODUCTION

For many years and even today, financial researchers and practitioners are convinced that the Capital Asset Pricing Model (CAPM) of Sharpe (1964) and Lintner (1965) constitutes an adequate instrument to explain return differences of financial assets, portfolios, funds, etc. Accordingly, it is the magnitude of a stock’s co-movement with the overall market, the systematic risk measure ‘Beta’, which determines a stock’s expected return. In practice, the CAPM serves as the most popular and common tool for the calculation of a firm’s cost of equity (Bruner, Eades, Harris & Higgins, 1998; Graham & Harvey, 2001; Brounen, De Jong & Koedijk, 2004), a variable, which is known to be important for financial managers and investors alike (Bartholdy & Peare, 2005). A survey conducted by Graham and Harvey (2001) shows that a large majority (73,5%) of the respondents make use of the CAPM in their decision-making processes. However, is one variable really able to explain stock return differences?

What is the role of general economic conditions or idiosyncratic risk factors of a firm? This thesis project aims to shed some more light on these questions by

extending as well as testing current asset-pricing models with a novel inflation exposure risk factor.

A company generally obtains capital from equity investors as well as lenders, in order to operate and to grow. In return, equity investors require compensation in the form of dividends or capital gains, whereas lenders are compensated by the payment of interest on the loan. From an equity capital provider’s point of view, an investment is only reasonable if the expected return on the investment is equal to or higher than the expected return on an alternative investment with the same quantity of risk. Similarly, financial managers are faced by a set of decisions, such as investment decisions, capital structure decisions or performance evaluation decisions; all of these require an appropriate estimation of the firm’s cost of equity as an input variable. More specifically, the cost of equity estimate serves as an input variable for the calculation of the firm’s cost of capital, most often the Weighted Average Cost of Capital (WACC), which is then used as a discount rate in a forward-looking capital budgeting project or as an input variable in the backward-looking performance measurement tool ‘Economic Value Added’ (EVA).

(5)

calculation of a firm’s cost of debt, which can often be easily observed by looking at the yield to maturity on the firm’s debt, the cost of equity estimation poses a great challenge to researcher. For instance, Bartholdy and Peare (2005) and Fama and French (1997) document the imprecision of the CAPM for the estimation of a firm’s cost of equity as well as an industry’s cost of equity, respectively. Similarly, numerous research projects refute the accuracy of this static one-period model and rather suggest an alternative to the classic CAPM, a multi-factor model, which includes additional factors aiming to explain asset returns more accurately. For instance, Ross (1976) proposes the so-called ‘Arbitrage Pricing Theory’ (APT), where a number of systematic risk factors affect stock returns. Despite the supposed superiority of such a model, academic research rather continued to support the theoretically derived CAPM, most probably because the APT does not identify and specify any additional factors (Hamberg, 2001). Similarly, Chen, Roll and Ross (1986) state that the co-movements of asset prices suggest the presence of macroeconomic risks. However, given the lack of theoretical foundations, they could not determine which economic variables are responsible. As a result of the ambiguity, Fama (1991) speaks of a “fishing license”.

In the early 1990’s, Fama and French (1992) proposed their so-called ‘three-factor model’ (FF3FM), which uses the original CAPM market ‘three-factor, but also takes into account two widely documented market anomalies, the so-called size-effect and value-effect. Subsequent research projects focused on a comparison between the CAPM and the FF3FM, whereby Fama and French (1992, 1993, 1996) as well as other researchers and practitioners claim that the CAPM is ‘dead’. Despite the alleged superiority of the FF3FM, tests of the model showed rather mixed results, especially when applied to out-of-sample populations (other than US market) and different underlying economic conditions. For instance, Hibbert and Lawrence (2010) found that the FF3FM performs marginally better than the CAPM, but to be comparable during bull/bear periods and Federal increasing/decreasing interest rate periods. Carhart (1997) also suggests an extended version of the FF3FM, because he claims that the FF3FM does not describe the stock return variation related to the continuation of short-run return patterns in the future (momentum effect). Thus, does the FF3FM

(6)

Throughout the last three decades a great number of research projects focused on the identification of relevant and significant economic risk factors. For instance, Robotti (2002) used a factor selection approach aiming at finding macroeconomic and financial market variables that are thought to capture the non-diversifiable risk of the economy. Amongst others, he found that inflation negatively affects average returns and commands a negative risk premium. From a logical point of view, inflation affects a company in numerous ways, all of them can have a negative or positive effect on the stock price. For instance, it may affect wages and the costs of supplies, the value of the assets that companies own, and the prices at which companies sell their products (Hillier, Westerfield, Jaffe & Jordan, 2010). Hence, depending on a company’s attributes, such as the type of business or pricing power, the ‘beta’ coefficient of inflation varies between companies. Accordingly, Duarte (2010) established that inflation exposure is a key determinant of risk in the cross-section of stocks, which cannot be explained by the Fama and French’s size factor or value factor. Similarly, Boons, De Roon and Szymanowska (2011) propose a novel inflation factor based on a stock’s ‘inflation-beta’, claiming to be the first to document an inflation risk premium that is non-negligible economically and statistically as well as robust across a wide range of specifications.

Based on the above outlined importance of a precise cost of equity estimation model, combined with the recent findings in regard to inflation exposure risk, this thesis project aims to answer the following underlying research question:

Is inflation exposure an important risk factor for the explanation of cross-sectional asset return variations in the Swedish and Dutch capital markets?

(7)

The objective of this thesis is twofold; it is of practical as well as theoretical nature. In regard to the former, it is my goal to provide a more sophisticated and precise model than the current CAPM and FF3FM by adding a novel inflation exposure risk factor, but at the same time it should be easy to use for investors and financial managers alike. Specifically, the suggested model should serve managers and investors for the cost of equity estimation and it should provide portfolio managers with insights regarding market anomalies, which can be used to adjust portfolio strategies as well as hedging strategies. In regard to the theoretical relevance, this thesis aims to close several academic literature gaps. First of all, I choose the Swedish and Dutch equity markets, since they have experienced comparatively little research in regard to asset pricing. Hence, I will provide further insights regarding the performance of the FF3FM in two out-of-sample populations as well as insights regarding the market efficiency and stock market anomalies for both markets. Second, this research project goes along with the novel work of Boons et al. (2011), thereby linking classic asset pricing literature with the most recent and up-to-date macroeconomic factor research. Last, asset-pricing models are generally tested on the basis of portfolio return data, however, in practice the models are mostly used to estimate the cost of equity for an individual firm or an industry. Hence, I gauge the success of the model on how well it explains the returns on single assets and industry sector portfolios.

(8)

markets suggesting that investors should adjust their investment strategies and hedging strategies according to the desired level of inflation exposure risk.

2. LITERATURE REVIEW

In order to satisfy the objectives of this thesis project, the theoretical framework touches upon different streams of academic research literature. More specifically, the Efficient Market Hypothesis (EMH) as well as the Modern Portfolio Theory (MPT) are shortly introduced, followed by a review of different asset-pricing models and the macroeconomic risk factor ‘inflation’. Based on the literature review, a hypothesis is developed which corresponds to the underlying research question.

(9)

value for the securities they issue. This underlying assumption is important throughout the thesis because a discovery of a market anomaly could either indicate some kind of market inefficiency or it could indicate that the equilibrium model (e.g. CAPM) might be less than a complete description of equilibrium price formation (Schwert, 2003; Keim, 2006). Both possibilities would be in favor of my suggested model.

Modern Portfolio Theory, introduced by Harry Markowitz in 1952, provides the groundwork for the popular Capital Asset Pricing Model. In short, Markowitz (1952) showed that investors can achieve a more efficient combination of risk and return if they focus on combinations of securities instead of individual securities alone. He found that a security’s unique (unsystematic) risk can be diversified away and thus leave the investor with the non-diversifiable (market or systematic) risk only. In other words, the variance of the return on a portfolio with many securities is more dependent on the covariance between the individual securities than on the variances of the individual securities (Hillier et al., 2010). Given Markowitz’s (1952) initial idea, Tobin (1958) made an improvement and added the Capital Market Line to the efficient frontier, arguing that the optimal investment decision is no longer a combination of securities only, but rather a combination of the tangency portfolio and the risk-free security. Despite the added complexity, all that really matters to measure the risk of a security is the responsiveness of a security to movements in the market portfolio, which is generally called an asset’s ‘Beta’.

The shortly described MPT constitutes the underlying framework for Sharpe

(1964) and Lintner’s (1965) famous Capital Asset Pricing Model (CAPM), which I have already mentioned in the previous section. So far, the Beta of a security is the appropriate risk measure in a diversified portfolio. Taking this finding as well as the assumption that investors are generally diversified, the CAPM states that the expected return on a security is positively related to its Beta. Hence, the CAPM is represented by the following equation:

!(!_!) = !_!+ !_! !(!_!) − !_! (1)

where !(!!) is the expected return on a security, !! is the risk-free rate, !! is the Beta of the security and !(!!) − !! is the difference between the expected return on the market and the risk-free rate of return.

(10)

market line (SML) provides a graphical illustration of Equation 1 and shows that the SML is upward sloping as long as theory and practice are correct that the expected return on the market E(R!) is greater than the return on risk-free assets (!!).

Moreover, the straight and upward sloping line illustrates that high-Beta securities should have a higher expected return than low-Beta stocks (Hamberg, 2001; Hillier et al., 2010). Last, Figure 1 shows that securities lying under or above the SML would be overpriced and underpriced, respectively. Arbitrage possibilities would exist, which investors could exploit by using homemade portfolios consisting of risk free-assets and the mispriced security. In equilibrium, every security must lie on the SML.

Figure 1 – Security Market Line

The above described SML relationship allows a test of the CAPM using the following excess return market model regression equation:

!_!!− !_!! = !_! + !_! !_!!− !_!! + !_!" (2)

where !!!− !!! is the excess return of the stock/portfolio ! over the risk-free rate !!!;!!!!−

!_!! is the excess return of the market portfolio over the risk-free rate; !_!! is the error term.

Amongst others, Jensen, Black and Scholes (1972) and Fama and MacBeth (1973) tested the CAPM for the 1930’s to the 1960’s and provide support by stating that high-Beta stocks are found to have higher average returns than low-Beta stocks. However, the documented relationship between Beta and average returns was not as steep as indicated by the theoretical SML. In contrast to the overall support for the

(11)

model, Roll (1977) argued that the CAPM is empirically untestable because the true market portfolio is not observable and stock indices or other measures of the market are poor proxies. Nevertheless, the CAPM was embraced with great popularity and numerous research projects extended the original CAPM in a variety of ways. The best known extensions refer to Merton’s (1973) Intertemporal-CAPM, which allows for multiple time periods and investment opportunities that change from one period to the next, or the International CAPM (Solnik, 1974; Stulz, 1981; Adler & Dumas, 1983), which is used to incorporate foreign exchange risk.

(12)

estimating asset returns, but also to use its potential for adjusting portfolio and hedging strategies.

Hence, similarly to the APT, Fama and French (1992) suggest that security returns are systematically affected by several factors, the market factor as well as the size factor and value factor. Thus, the relationship between risk and return can be expressed as:

!!!− !!! = !! + !! !!!− !!! + !!!"#!+ ℎ!!"#!+ !!! (3)

where !!!− !!! is the excess return of the stock/portfolio ! over the risk-free rate !!!;!!!!−

!_!! is the excess return of the market portfolio over the risk-free rate;!!"#! and !"#!!are

the two Fama and French variables representing the size factor and value factor, respectively; !!! is the error term.

However, the FF3FM lacks a theoretical foundation and is classified as an empirically based asset-pricing model (Hamberg, 2001). That means, the two added factors are regarded as market anomalies, which are defined as empirical results that seem to be inconsistent with maintained theories of asset pricing behavior (Schwert, 2003). As already outlined in a previous section, Schwert (2003) also claims that anomalies indicate either some form of market inefficiency (profit opportunities) or inadequacies in the underlying asset-pricing model. However, in order to prevent ‘data snooping’ (Lo & MacKinley, 1990) and to provide some kind of confirmation for the existence of a size effect or value effect, those anomalies must be observable across different sample periods and sample populations. In recent literature, tests of the FF3FM showed rather mixed results. For instance, Vosilov and Bergstroem (2010) and Kilsgard and Wittorf (2010) found contradicting results regarding the size effect and value effect for the Swedish capital market. Hibbert and Lawrence (2010) tested the CAPM and FF3FM using individual stock returns and found that the FF3FM performs marginally better than the CAPM. Given an active discussion of the performance of the different asset-pricing models, the FF3FM clearly provides a good starting point for further research in the field of asset pricing.

(13)

(14)

and stock returns. In subsequent research projects, Geske and Roll (1983), Stulz (1986) and Marshall (1992) also aimed at finding explanations for the ‘stock return – inflation puzzle’; however, with diverse approaches and results. Notably, Geske and Roll (1983) state that the following fiscal and monetary linkage from stock returns to money base growth is firmly in place:

… a random negative (positive) real shock affects stock returns which, in turn, signal higher (lower) unemployment and lower (higher) corporate earnings. This leads to lower (higher) personal and corporate tax revenues. Government expenditures do not change to accommodate the change in revenues so the Treasury's deficit increases (decreases). The Treasury responds by increasing (decreasing) borrowing from the public. The Federal Reserve System purchases some of the change in Treasury debt and eventually pays for it by expanding (contracting) the growth rate of base money. Higher (lower) inflation is induced by the altered money base growth rate. Rational investors realize that a random real shock signaled by the stock market will trigger this chain of fiscal and monetary responses. Thus, they alter the prices of short-term securities contemporaneously with the stock return signal. To the extent that an increased (decreased) deficit, triggered by a real shock, is not expected to be "monetized" by the Federal Reserve, the Treasury's increased (decreased) supply of debt securities can also cause an increase (de- crease) in real interest rates. Investors decide collectively on whether a particular stock return signifies a change in real rates, in expected inflation rates, or in both. Regardless of the mix between real rate and expected inflation, nominal interest rates must change. (p. 28 f.)

(15)

behavior of aggregate stock market indices co-varies with inflation, leaving out the behavior of individual stocks. In this respect, Ang et al. (2012) refer to the considerable heterogeneity across companies and their respective differences in pricing power, which is a company’s ability to set prices or to pass through increased input prices. By focusing on individual stock returns, Ang et al. (2012) documented substantial variation in how individual stocks co-vary with inflation, whereas the correlation of the aggregate market with inflation is found to be negative. Thus, they conclude that the overall stock market may be a poor inflation hedge, but companies in certain sectors or with certain characteristics may have better inflation-hedging properties than other companies. Furthermore, Ang et al. (2012) document that stocks with a history of being good inflation hedges have high nominal and real returns, on average. Boons et al. (2011) build upon the work of Duarte (2010) and Ang et al. (2012) and claim to be the first to document an inflation risk premium that is non-negligible, economically and statistically. Leaving open the question why firms’ inflation exposures differ in the first place, they make use of Fama and French’s portfolio approach and include an additional factor, which aims to capture the common variation in returns related to inflation-betas. More specifically, Boons et al. (2011) suggest a zero-investment portfolio (low inflation-beta stocks minus high inflation-beta stocks), which is able to capture the outperformance (by 8% per year) of low inflation-beta stocks in the period from 1964 to 2003 and the underperformance (by -12% per year) of low inflation-beta stocks in the post-2003 period. This reversal of the correlation between inflation and the stock market around the millennium is also documented by Bekaert and Wang (2010) and Campbell et al. (2011).

(16)

H1: The revised FF3FM, including an additional inflation exposure risk factor, does not explain the cross-sectional return variations of individual stocks or industry sector portfolios in the Swedish and Dutch capital markets.

The corresponding factor model (revised FF3FM) is represented by the following equation:

!_!!− !_!! = !_! + !_! !_!!− !_!! + !_!!"#_!+ ℎ_!!"#_!+ !_!!"#$_!+ !_!! (4) where !!!− !!! is the excess return of the stock/portfolio ! over the risk-free rate !!!;!!!!−

!_!! is the excess return of the market portfolio over the risk-free rate; !"#! and !"#!!are

the two Fama and French variables representing the size factor and value factor, respectively; !"#$_! is the added inflation exposure risk factor and !!! is the error term. This research project differentiates from previous research by combining the latest findings in regard to inflation risk with the commonly known Fama and French approach. To my knowledge, the work of Boons et al. (2011) is the only other research project to this point with a similar approach, however, focusing on a different sample population, different sample period and portfolio returns instead of individual stock returns or industry sector portfolio returns. Moreover, I have adjusted the construction of the inflation exposure risk factor to the documented reversal of the correlation between inflation and the stock market (the updated working paper by Boons et al. (18th of November 2012) also made an adjustment to it). I hope to achieve valuable results in regard to the novel inflation exposure factor, thereby providing support for a more accurate cost of equity estimation model as well as obtaining insights regarding the presence of market anomalies. Finally, this thesis project investigates the Swedish and the Dutch capital markets because they have only experienced little research in regard to the FF3FM, not to mention in regard to inflation risk. Furthermore, addressing two independent samples adds to the robustness of the results and can also be used in order to devise hedging strategies or to make use of arbitrage opportunities on an international level.

3. METHODOLOGY

(17)

3.1 Data Selection

The data described in the following sections is obtained from the DATASTREAM database. Information regarding the annual index composition of the OMX Stockholm Benchmark index and the AEX25 index and the AMX25 index are retrieved from DATASTREAM and NYSE EURONEXT, respectively. Relevant academic literature is obtained through the search engines of the libraries of the University of Groningen and Uppsala University, especially the electronic databases ‘Business Source Premier’ and ‘Science Direct’.

Sample period. The sample period stretches over a time period of 6 years and 8

months, from the 1st of January 2006 to the 1st of September 2012. In general, the choice for this sample period relates to the availability of stock return data for the NASDAQ OMX Nordic Stockholm stock exchange on the DATASTREAM database. Furthermore, the choice of the sample period takes into account the findings in regard to time-varying Betas. Groenewold and Fraser (2000) propose a superiority of a six- and seven-year rule of thumb for the Beta calculation over the widely used five-year rule of thumb. Hence, my sample period selection is in line with these findings, which assumingly adds to the validity of the results.

Sample. This research project investigates the Swedish and Dutch capital markets

(18)

an annual basis, a stock had to be continuously listed from the first index review in year t to the first index review in year t+1 in order to be included in the sample. The NASDAQ OMX Stockholm Benchmark index is reviewed at the beginning of each year, whereas the AEX25 index and the AMX25 index are reviewed according to the actual review dates (in March year t). As an example, a stock that entered the index at the first index review in March 2010 and was deleted again at the second index review in September 2010 is excluded from the sample. In contrast, if the stock had been deleted during the first index review in 2011, the stock would still have been included for the sample in year 2010. Following this approach assures not to exclude failing companies from this study because they no longer exist or were delisted for some other reason, thereby preventing survivorship bias. Last, although monthly return data is used, the use of an all-share index, which also contains small-size firms, would increase the possibility of non-trading bias.

3.2 Data Description

The construction of the dependent and explanatory variables demands several different types of data and follows Fama and French (1992). Each data type is explained in the following:

Individual stock return (!_!). Monthly log return data for each individual stock of the above described samples. The return index (RI) data type is obtained, which assumes that dividends are re-invested to purchase additional units of equity at the closing price applicable on the ex-dividend date (DATASTREAM).

Return on market (!_!). Monthly log return data for the OMXSB index and the AEX all-share index. The OMXSB index is used because OMXS all-share index return data

is only available from 1st of January 2008. However, the OMXSB index is assumed to

be a very good performance indicator of the NYSE OMX Stockholm stock exchange. Again, the return index (RI) data type is obtained, where dividends are re-invested. Both indices are constructed as value-weighted portfolio returns (DATASTREAM).

Risk-free interest rate (!!). The choice follows the recommendation of Thomson

(19)

bill is used for the Swedish market and the 3-month interbank rate (EURIBOR) is used for the Dutch market. The monthly log returns are calculated based on the obtained return index (RI) data type (DATASTREAM).

Book-to-Market equity ratio (BE/ME). Fama and French (1992) define

book-to-market equity as book common equity for the fiscal year ending in calendar year t-1, divided by market equity at the end of December of year t-1. I am following this approach by taking the reciprocal of the obtained Price-to-Book value (PTBV) data from DATASTREAM. It is tested and assumed that the annual PTBV data from the

1st of January in calendar year t includes the fiscal year end book common equity of

year t-1 and therefore is an appropriate proxy for Fama and French’s BE/ME. According to Thomson Reuters’ explanation, the book value per share used in the PTBV calculation represents the book value at the company’s fiscal year end for non-U.S. corporations. Hence, annual PTBV data retrieved in June of calendar year t would be misleading since the ME data will also be based on the value in June year t.

Market capitalization (ME). The market capitalization is calculated as the share price

multiplied by the number of ordinary shares in issue. In order to follow Fama and French (1992), I obtain the Market Value (MV) data type for year t at the end of December year t-1 (DATASTREAM).

Inflation. The Swedish and Dutch Consumer Price Indices (CPI) constitute the

inflation proxies for the respective markets. Since the data is already obtained in percentage format, the simple monthly difference is calculated (DATASTREAM).

3.3 Construction of the variables

Dependent variables. In contrast to Fama and French (1992) and most other research

(20)

available. As a result, a total of 43 Dutch stocks and 56 Swedish stocks are analyzed. See Appendix A and B for the complete constituent lists. The final dependent variable of each selected stock is the monthly excess return over the risk-free interest rate. The industry portfolio returns are calculated as monthly value-weighted log returns based on the ICB industry classification model (level 2 sector). Thus, the previously selected stocks are assigned to its respective industry sector: Consumer Services, Financials, Telecommunications, Technology, Industrials, Consumer Goods, and Oil & Gas. In addition, the Swedish market contains stocks in the ‘Healthcare’ as well as the ‘Basic Materials’ industry sectors. See Appendix A and B for the respective constituent lists.

Fama and French factors. The explanatory variables are constructed in accordance

with Fama and French’s (1992) approach. Therefore, I sort all stocks of the Swedish sample and the Dutch sample in regard to its ME (size) and BE/ME (value). The stocks are sorted into a total of two size groups with the median as a cut-off point (Figure 2) as well as three BE/ME groups based on the breakpoints for the bottom 30% (Low), middle 40% (Medium) and top 30% (High) (Figure 3). The decision to form three BE/ME groups and only two size groups also follows Fama and French (1992; 1993), where they have found evidence that BE/ME has a stronger role in the explanation of average stock returns than size has.

Figure 2 – Size Sorting

Figure 3 – BE/ME Sorting Median

Size

Small Big

Figure 2. Size sorting, where the stock belongs to the big-size group.

Bottom 30%

BE/ME

Low Medium High

Top 30%

(21)

From the intersections of the two size groups and three BE/ME groups, six portfolios (S/L, S/M, S/H, B/S, B/M, B/H) are formed. Since the sorting of the size and BE/ME groups is performed separately of each other, each stock is assigned to one size group and one BE/ME group. Figure 4 illustrates the sorting procedure, where the stock belongs to the big-size group as well as the low-BE/ME group. Next, monthly value-weighted returns on the six portfolios are calculated from the beginning of January of year t to the beginning of January t+1. Making use of value-weighted portfolio returns has the advantage of minimizing variance and to capture the return behaviors of realistic investment opportunities (Fama & French, 1993).

Figure 4 – Double Sorting

To be included in the ME and BE/ME grouping in year t, monthly stock return data for year t as well as the stock’s ME of year t-1 and BE/ME for year t-1 must be available. As described earlier, ME is the market equity at the end of December of year t-1 and BE/ME is the fiscal year end book common equity of calendar year t-1, divided by the market equity at the end of December year t-1.

In order to avoid survivorship bias, the actual index constituent lists of year t are operationalized. In contrast to Fama and French’s (1992) ranking in June each year, I conduct the ranking at the beginning of January year t and retrieve the respective monthly stock returns from 1st of January year t to 1st of January year t+1. A June ranking is not appropriate since the fiscal-year end BE-data for t-1, which is

70% Big BE/ME Size Low! 30% Small Medium! High Median

(22)

necessary for the BE/ME calculation, is already available via DATASTREAM at the beginning of the respective year t.

SMB. In order to account for the ‘size effect’, the zero-investment portfolio SMB (small minus big) is constructed. The portfolio takes a long position in small-cap stocks and a short position in big-cap stocks, whereby it aims to mimic the risk factor in stock returns related to size. More specifically, the SMB portfolio is calculated as the monthly difference between the simple average of the returns of the three small-size portfolios (S/L, S/M, S/H) and the simple average of the returns of the three big-size portfolios (B/L, B/M, B/H). Following this approach assures that the SMB portfolio is not influenced by BE/ME (Fama and French, 1993).

!"#_!= (!!"#$$%&',!!!!"#$$%&'()",!!!!"#$$%&'!,!)

! −

(!!"#$%&,!!!!"#$%&"'(,!!!!"#$"#!,!)

!

HML. Similarly, the HML (high minus low) variable is constructed as a

zero-investment portfolio aiming to mimic the risk factor in stock returns related to BE/ME. Thus, the HML portfolio takes a long position in high-BE/ME stocks (‘value’ stocks) and a short position in low-BE/ME stocks (‘growth’ stocks). In contrast to the SMB variable, HML is calculated as the monthly difference between the simple average of the returns on the two high-BE/ME portfolios (S/H, B/H) and the simple average of the returns on the two low-BE/ME portfolios (S/L, S/H). As a result, the HML portfolio returns should be largely free of the influence of the size factor in returns (Fama & French, 1993)

!"#_! =(!!"#$$%&'!,!!!!"#$"#!,!)

! −

!!"#$$%&',!!!!"#$%&,!

!

!_!− !_!.!The Swedish and Dutch market returns are proxied by the OMXSB index

return and the AEX all-share index return, respectively. Hence, both indices are value-weighted market portfolio returns. The market return variable is calculated as the monthly excess return on the market portfolio over the risk free interest rate.

HILO. The inflation exposure risk factor, which I named ‘HILO’, is constructed by

(23)

log returns of each stock are regressed against the monthly change in the inflation rate in order to obtain each stock’s inflation-beta. Thereby the sample period is divided into two sub-periods with a cut-off point in January 2008. This choice mainly relates to the recent financial crisis and the correspondingly suspected change of betas. Next, the stocks are grouped into one of three portfolios based on the inflation-beta breakpoints for the bottom 30% (Low), middle 40% (Medium) and top 30% (High). Afterwards, monthly value-weighted returns are calculated for each portfolio. Finally, the zero-investment portfolio HILO is calculated by subtracting the low inflation-beta portfolio from the high-inflation beta portfolio. As a result, the HILO factor should be able to capture the common variation in stock returns related to inflation exposure risk.

!"#$_! = !_{!!"!!!"#$%&!'"!!"#$,!}− !_{!"#!!"#$%&!'"!!"#$,!}

3.4 Estimation Method

The hypothesis is tested by operationalizing the revised FF3FM (incl. the inflation exposure risk factor ‘HILO’) for each selected stock as well as the value-weighted industry sector portfolios of the two capital markets for the specified sample period.

Revised FF3FM:

!_!"− !_!" = !_! + !_! !_!"− !_!" + !_! !"#_! + ℎ_! !"#_! + !_!(!"#$_!) + !_!" (4) where !!"− !!" is the excess return of the stock/value-weighted industry portfolio ! over the risk-free rate !_!";!!_!"− !_!" is the excess return of the market portfolio over the risk-free rate; !"#! and !"#!!are the two Fama and French variables representing the size factor and value factor, respectively; !"#$_!represents the inflation exposure risk factor; !!" is the error term.

(24)

AutoRegressive Conditional Heteroskedasticity. In this respect, problems concerning heteroskedasticity are often found in stock return data, which could lead to inefficiencies of the regression model. Accordingly, the results of the preliminary uncorrected OLS regressions show significant White test results in approximately 40% and 30% of the cases for individual stocks in the Swedish market and Dutch market, respectively (Table 1). In contrast, the Arch test only shows significant results in a very few cases. Thus, the final estimation method is an OLS regression with White heteroskedasticity-consistent standard errors & covariance. However, in the presence of Arch effects, a GARCH(1,1) model is estimated in order to check if it leads to different results.

Table 1 - Preliminary uncorrected OLS White test results

Individual stocks Industry portfolios

Sweden 23/41% 4/44%

The Netherlands 13/30% 4/50%

Table 1 reports the preliminary uncorrected OLS White test results - Number and proportion of significant cases at the 5 % significance level.

In regard to the hypothesis test, the above described regression analysis provides the coefficient estimates and the respective significance levels from the t-tests. A statistically significant HILO coefficient will provide a first indication regarding the applicability of inflation exposure risk for the respective stock or industry sector portfolio. Moreover, the F-Statistic is looked at, but not documented, and provides a first indication of the overall statistical significance of the model. Furthermore, the adjusted !!_{is documented as an indication of the explanatory power of the}

(25)

variables test is performed in order to test the necessity of the HILO factor. This means, the HILO variable is dropped from the regression equation, whereby the revised FF3FM is formally tested against the original FF3FM, at least in respect to the suggested HILO factor. Conducting the above-described analysis will provide the necessary information in order to reject or to accept the hypothesis. This will ultimately provide a sophisticated answer to the research question. All regressions were conducted using the statistical package EViews 7, whereas the data was prepared in Microsoft Office Excel.

4. EMPIRICAL RESULTS & ANALYSIS

This chapter contains the general summary statistics of the independent and dependent variables and more importantly the empirical results of the previously described estimation method. The results are presented in conjunction with a brief analytical discussion.

In advance to reporting the actual regression results, the summary statistics of the previously described data provide a first impression of the presence of market anomalies in the Swedish and Dutch markets. Given Fama and French’s (1992) argumentation in regard to the size effect and the value effect, as well as the recently documented inflation–beta effect (Duarte, 2010; Boons et al., 2011; Ang et al., 2012), these suggested anomalies should be directly observable through the value-weighted returns on the explanatory variables SMB, HML and HILO, respectively. That means, the size effect would be confirmed by an average positive return on the SMB portfolio. Put differently, it would indicate that, on average, small-size companies have outperformed big-size companies in the years 2006 to 2012. The same applies to the HML and HILO variables.

(26)

the presence of the value effect in the Swedish market and the suggested inflation-beta effect in both markets. Put differently, high-BE/ME stocks outperformed low-BE/ME stocks in the Swedish market by an average of 0,304% per year during the period 2006 to 2012. Furthermore, high inflation-beta stocks outperformed low inflation-beta stocks by an annual average of 0,766% and 1,568% in the Swedish and Dutch markets, respectively. In regard to the mixed results of the value effect and size effect, Schwert’s (2003) notion might provide an explanation:

“… possibility that anomalies are more apparent than real. The notoriety associated with the findings of unusual evidence tempts authors to further investigate puzzling anomalies and later to try to explain them. But even if the anomalies existed in the sample period in which they were first identified, the activities of practitioners who implement strategies to take advantage of anomalous behavior can cause the anomalies to disappear (as research findings cause the market to become more efficient)”. (p. 941)

Similarly, Keim (2006) also suggests that there is no guarantee for market anomalies to persist in the future, even if some effects could be observed for nearly 100 years. Accordingly, I propose that the novelty of the inflation-beta effect, and the corresponding nescience of practitioners, did not lead to the implementation of appropriate investment strategies yet. As a result, the inflation-beta effect is still observable. However, statistical tests, such as an Independent Samples Test, are required to provide further insights regarding the statistical significance of the differences in means. The argumentation behind the market anomaly itself is beyond the scope of this research project and definitely needs to be investigated in depth by future research.

(27)

Table 2: Summary statistics of monthly return data: Sweden

Panel A – Market Data (Sweden)

Mean Maximum Minimum StdDev.

!_! 0,153% 6,579% -7,244% 2,584%

!_! 0,068% 0,158% 0,002% 0,05%

SMB -0,149% 4,573% -5,464% 1,887%

HML _-0,004% _7,386% _-8,993% _2,826%

HILO -0,044% 6,242% -12,514% 2,974%

Panel A and Panel B show the summary return statistics for the independent and dependent variables, respectively. The first row considers the monthly data, whereas the second row reports the average annual return for each variable. The return statistics are based on the sample period 01.2006 – 09.2012.

Panel B – Individual Stocks & Industry Portfolios (Sweden) Mean Maximum Minimum StdDev.

N=56 stocks 0,266% _3,752% 38,512% -23,577% 4,446% Consumer Services 0,414% _4,874% 6,376% -7,184% 2,563% Financials 0,049% 0,937% 9,636% -12,069% 3,431% Telecommunications 0,303% 3,711% 7,779% -8,476% 2,718% Technology -0,329% -4,036% 11,578% -13,531% 3,875% Industrials 0,262% 3,568% 8,084% -10,218% 3,449% Consumer Goods 0,263% _3,507% 8,251% -5,558% 2,484%

Oil & Gas 0,102% _1,074% 13,143% -15,976% 4,835%

Healthcare 0,065% _0,585% 4,702% -4,597% 2,060%

(28)

Table 3: Summary statistics of monthly return data: The Netherlands

Panel A – Market Data (Netherlands)

Mean Maximum Minimum StdDev.

!_! 0,011% 5,478% -8,835% 2,766%

!! 0,088% 0,203% 0,013% 0,059%

SMB -0,050% 4,110% -4,119% 1,561%

HML -0,176% 7,716% -7,107% 2,639%

HILO -0,893% 7,514% -21,890% 3,782%

Panel A and Panel B show the summary return statistics for the independent and dependent variables, respectively. The first row considers the monthly data, whereas the second row reports the average annual return for each variable. The return statistics are based on the sample period 01.2006 – 09.2012.

Panel B – Individual Stocks & Industry Portfolios (Netherlands) Mean Maximum Minimum StdDev.

N=43 stocks -0,089% _-0,121% 22,050% -35,655% 4,780% Consumer Services -0,009% 0,073% 4,067% -6,265% 2,223% Financials 0,926% 12,277% 21,181% -33,578% 8,126% Technology 0,113% _3,187% 7,383% -13,490% 4,433% Industrials -0,109% _-0,441% 7,670% -9,843% 3,303% Consumer Goods 0,312% _3,814% 4,961% -6,516% 2,228%

Oil & Gas 0,203% _2,356% 6,530% -7,745% 2,759%

(29)

4.1 Individual stock returns

Panel A of Table 4 reports the aggregate regression results of the revised FF3FM (including the inflation exposure risk factor) for the individual stocks and Panel B provides similar results for the industry portfolios of the two capital markets. In particular, Table 4 presents the coefficient estimates, the respective statistical significance, as well as the adjusted R!_{. In order to test the hypothesis, I empathize the}

results regarding the model’s goodness rather than focusing on the interpretation of the individual coefficient estimates. Accordingly, the regression analyses using individual stock returns illustrate that the intercept ! is statistically significant at the 5% level in only 2 cases (3,6%) for the Swedish market and 5 cases (11,6%) for the

Dutch market. A positively statistical significant ! means that the model

underestimates the returns for these stocks, whereas a negatively statistical significant ! indicates that the model overestimates the returns for these stocks. The average ! estimate is 0,0014 and -0,0004 for Sweden and the Netherlands, respectively. The combination of these findings shows that the revised FF3FM is doing well in estimating individual stock returns, but given the amount and proportion of significant !’s, the model performs better for individual stock returns in the Swedish market than the Dutch market. These results are also supported when I estimated a GARCH(1,1) model for the cases, which indicated Arch effects. In addition to the results of the intercept !, the adjusted R!_{is reported to measure the model’s level of explanation. In}

both markets, the adjusted R!_{!ranges from less than 10% to approximately 80% and}

has an average adjusted R!_{of approximately 50%. In order to put the results into}

perspective, Table 5 provides a comparison between the regression results of the original FF3FM and the revised FF3FM. This comparison shows that the average

adjusted R!_{of the revised FF3FM is marginally higher than for the original FF3FM:}

1,73% (Sweden) and 2,78% (The Netherlands). It is notable that a previous research

project by Hibbert and Lawrence (2010) reports an even lower average R!_{(27%) for}

the original FF3FM in regard to individual stock returns traded on the NYSE over the period 1963 to 2006. However, my results do not reach the reported average R!_of

(30)

Finally, Table 6 reports the results of the redundant variables test, whereby the revised FF3FM’s specification is further analyzed and contrasted against the original FF3FM. The results of this test show that the additional inflation exposure risk factor ‘HILO’ is not considered as redundant at the 5% significance level in 39,5% and 37,5% of the cases in the Swedish and Dutch capital markets, respectively. Similarly, Panel A of Table 4 reports that the HILO coefficient estimate is significant at the 5% level in 35,7% of the Swedish stocks and 32,6% of the Dutch stocks. Thus, the results of the redundant variables test (Table 6) as well as the results of the regression analysis (Table 4) indicate that the HILO factor is important for a non-negligible percentage of the cases in both markets. It also shows that the HILO coefficient estimate is significant in more cases than Fama and French’s SMB factor and HML

factor for the Swedish market. However, unlike the CAPM factor (!_!− !_!), which is

significant in all cases, the HILO factor is not important to all stocks. In the Dutch market, the number of significant coefficient estimates of the HILO factor is not as convincing, but still comparable to Fama and French’s HML factor (Table 4).

Based on the described regression results, in particular the intercept ! and the redundant variables test, I find that the revised FF3FM is an appropriate and well-specified model for the explanation of individual stock returns. Although the comparison with the original FF3FM only shows a marginal improvement of the average adjusted R!_{and no significant difference of the intercept results, the}

redundant variables test shows that the HILO factor is significant. Moreover, the regression results suggest that the HILO factor is significant in more cases than the SMB factor and the HML factor in the Swedish market and has a similar importance as the HML factor in the Dutch market. Hence, I confirm the findings of Boons et al. (2011, 2012) and conclude that an extra inflation exposure factor improves the quality of current asset-pricing models to explain cross-sectional return variations of individual stocks.

4.2 Industry sector portfolio returns

(31)

Table 4. Regression results: Average Coefficient Estimates & Number/Proportion of stocks/industry portfolios with significant parameters.

(OLS with White heteroskedasticity-consistent standard errors & covariance) Panel A – Individual Stocks

! !!-!! SMB HML HILO Adjusted !! Max !! Min !!

Sweden Average coefficient 0,0014 1,0519 0,2026 0,0341 0,1055 49,65% 86,87% 5,84% 5% 2 / 3,6% 56 / 100% 17 / 30,4% 10 / 17,9% 20 / 35,7% 10% 4 / 7,1% 56 / 100% 24 / 42,9% 12 / 21,4% 24 / 42,9% The Netherlands Average coefficient -0,0004 0,9866 0,4680 0,1224 0,0178 50,29% 79,64% 8,84% 5% 5 / 11,6% 39 / 90,7% 27 /62,8% 14 / 32,6% 14 / 32,6% 10% 9 / 20,9% 40 / 93% 28 / 65,1% 16 / 37,2% 17 / 39,5%

This table (Panel A and Panel B) reports the regression results for the following model:

!_!"− !_!" = !_!+ !_! !_!"− !_!" + !_! !"#_! + ℎ_! !"#_! + !_!(!"#$_!) + !_!" (Revised FF3FM)

where !!"− !!" is the excess return of the stock ! or the value-weighted industry portfolio returns over the risk-free rate !!";!!!"− !!" is the excess return of the market portfolio over the risk-free rate; !"#! and !"#!!are the two Fama and French variables representing the size factor and value factor, respectively; !"#$!!represents the inflation exposure risk factor, !! is the intercept and !!" is the error term.

Panel A provides the regression results in regard to individual stock returns (aggregate), whereas Panel B (next page) shows the regression results in regard to the industry portfolio returns. It includes the average coefficient estimate, the average adjusted !!_{, maximum}_!!_{and minimum}_!!_{. Furthermore, the two} rows below the average coefficient estimates show the number and proportion of stocks/portfolios with significant parameters at the 5% and 10% significance level, respectively. Results at the 10% significance level include the results from the 5% significance level.

(32)

Table 4 (continued). Panel B – Industry Portfolios

Sweden ! !_!-!_! SMB HML HILO Adj.!!

Consumer Services 0,0028 *0,8086 0,0675 *-0,3469 *-0,2043 56,52% Financials -0,0013 *1,1892 -0,0468 *0,2373 -0,0847 83,83% Telecom. 0,0019 *0,5719 -0,0450 0,0562 0,1281 37,14% Technology -0,0055 *0,9159 -0,4140 -0,2049 -0,2259 35,47% Industrials 0,0010 *1,1550 0,0020 -0,0038 *0,2029 91,54% Consumer Goods 0,0014 *0,8379 0,1149 0,0444 -0,0763 67,37%

Oil & Gas 0,0004 *0,8086 0,3598 *0,3624 *0,4225 38,32%

Healthcare -0,0004 *0,4587 0,1008 -0,0925 *-0,3646 27,17%

Basic Materials -0,0034 *1,3936 0,1766 0,1349 *0,2486 73,59%

Average coefficient -0,0004 0,9044 0,0351 0,0208 0,0051 56,77%

5% 0/ 0% 9/100% 0/0% 3/33,3% 5/55,6%

10% 0/0% 9/100% 1/11,1% 3/33,3% 6/ 66,7%

The Netherlands ! !_!-!_! SMB HML HILO Adj. !!

Consumer Services -0,0017 *0,7548 -0,0047 0,0145 *-0,1456 62,94%

Financials *0,0149 *1,8026 *1,0800 *0,8706 0,3411 75,94%

Technology -0,0009 *1,5470 -0,0412 -0,1218 -0,2326 65,85%

Industrials -0,0015 *1,1383 0,0293 0,0350 -0,0526 84,76%

Consumer Goods 0,0000 *0,8768 0,0648 0,0340 *-0,3337 71,04%

Oil & Gas -0,0006 *0,9943 -0,1784 *-0,2073 *-0,2265 63,23%

Basic Materials 0,0026 *1,1608 0,0495 -0,0847 *0,5236 90,47%

Average coefficient 0,0018 1,1789 0,1428 0,0772 -0,0180 73,46%

5% 1/14,3% 7/100% 1/14,3% 2/28,6% 4/57,1%

10% 1/14,3% 7/100% 1/14,3% 2/28,6% 5/71,4%

sectors, the previous findings from the individual stock returns are carried on to the industry portfolio returns. Thus, the regression results propose that the intercept is close to zero and the model is accurately specified. The intercepts of the industry portfolio regressions average -0,0004 for the Swedish market and 0,0018 for the Dutch market. These results are also supported when I estimated a GARCH(1,1) model for the cases, which indicated Arch effects. The adjusted R!_{ranges from}

27,17% to 91,54% for the Swedish industry sector portfolios and from 62,94% to

90,47% for the Dutch industry sector portfolios. The average adjusted R!_{amounts to}

(33)

Table 5. FF3FM vs. revised FF3FM – Regression Results

(OLS with White heteroskedasticity-consistent standard errors & covariance) Individual Stocks Adjusted _!_! Max _!_! Min _!_! ! !_!!

-! SMB HML HILO Sweden FF3FM 47,92% 84,86% 1,03% _3,6%2 _100%56 _32,1%18 _16,1%9 Revised FF3FM 49,65% 86,87% 5,84% 2 3,6% 56 100% 17 30,4% 10 17,9% 20 35,7% The Netherlands FF3FM 47,51% 79,09% 8,79% 4 9,3% 42 97,7% 23 53,5% 13 30,2% Revised FF3FM 50,29% 79,64% 8,84% 5 11,6% 39 90,7% 27 62,8% 14 32,6% 14 32,6% Industry Portfolios Adjusted _!_! Max _!_! Min _!_! ! !_!!

-! SMB HML HILO Sweden FF3FM 52,78% 89,23% 6,25% _0%0 _100%9 _0%0 _22,2%2 Revised FF3FM 56,77% 91,54% 27,17% 0% 0 100% 9 0% 0 33,3% 3 55,6% 5 The Netherlands FF3FM 68,78% 84,78% 55,59% 1 14,3% 7 100% 2 28,6% 2 28,6% Revised FF3FM 73,46% 90,47% 62,94% 1 14,3% 7 100% 14,3% 1 28,6% 2 57,1% 4 Table 5 reports the regression results for the following two models:

FF3FM

!!"− !!" = !! + !! !!"− !!" + !! !"#! + ℎ! !"#! + !!"

Revised FF3FM

!!"− !!" = !! + !! !!"− !!" + !! !"#! + ℎ! !"#! + !!(!"#$!) + !!"

where !!"− !!" is the excess return of the stock ! or the value-weighted industry portfolio returns over the risk-free rate !_!";!!_!"− !_!" is the excess return of the market portfolio over the risk-free rate; !"#! and !"#!!are the two Fama and French variables representing the size factor and value factor, respectively; !"#$!!represents the inflation risk factor, !! is the intercept and !_!" is the error term. Table 5 presents the average adjusted !!_{, maximum}_!! and minimum !!_{. Furthermore, the number and proportion of stocks/portfolios with} significant parameters at the 5% significance level are reported.

(34)

(57,1%), whereby the SMB and HML coefficient estimates are significant in a lower percentage of cases (0% - 33,3%).

Last, the results of the redundant variables test (Table 6) provide substantial support for the inclusion of the HILO factor in the model. More specifically, 5 out of the 9 Swedish industry sectors (55,6%) and 6 out of the 7 Dutch industry sectors (86,7%) do not consider the HILO factor as redundant at the 5% significance level. Furthermore, the reported results of the regression analysis (Table 4) show statistically significant coefficient estimates of the HILO factor for 5 Swedish industry portfolios (55,6%) and 4 Dutch industry portfolios (57,1%), whereby the SMB and HML coefficient estimates are significant in a lower percentage of cases (0% - 33,3%).

Table 6. Redundant Variables Test - HILO

5% 10%

Sweden Individual Stocks 21 / 37,5% 26 / 46,4%

Industry Portfolios 5 / 55,6% 6 / 66,7%

The Netherlands

Individual Stocks 17 / 39,5% 18 / 41,9%

Industry Portfolios 6 / 85,7% 6 / 85,7%

Number/proportion of Redundant Variable Test results where p-value <0,05 and p-value <0,10, respectively.

Similarly to the analysis results of the revised FF3FM using individual stocks return data, the results of the regression analysis in regard to industry sector portfolio returns propose that the model is accurately specified and the redundant variables test provides strong support for the inclusion of the HILO factor in both capital markets. Moreover, the comparison with the regression results from the original FF3FM further indicates the superiority of the revised FF3FM (Table 5).

(35)

industry sector portfolio returns than for individual stock returns. The findings confirm and extent previous findings from Duarte (2010), Boons et al. (2011) and Ang et al. (2012), that an asset’s inflation-beta should be regarded as a reasonable risk factor for the pricing of financial assets. Since a detailed analysis and explanation of the previously mentioned ‘stock return – inflation puzzle’ is beyond the scope of this research project, a further investigation regarding the reasons for the heterogeneity of inflation-betas and its corresponding effect on stock returns is needed. Moreover, despite the overall approval of the additional HILO factor, a more detailed analysis of its coefficient estimates is left out in this research project and should be addressed by future research, thereby sorting out the relationship between an asset’s inflation-beta and the respective return. It is also worth mentioning that the suggested Fama and French factors were less significant in explaining return differences than the introduced HILO factor. This might again be explained by Schwert’s (2003) notion that market anomalies often cease to exist after they have been the focus of academic research and practitioners adjusted their investment strategies accordingly.

To sum up, this research project provides evidence for the presence of an inflation-beta effect in the Swedish and Dutch capital markets, but could not provide convincing proof for the existence of a size effect or value effect. Moreover, the revised FF3FM, including the added inflation-beta risk factor ‘HILO’, is found to be an appropriate model to explain individual stock returns as well as industry portfolio returns in the Swedish and Dutch capital markets. Thus, financial managers and investors should take an asset’s inflation-beta into account for the calculation of a company’s cost of equity or an industry’s cost of equity. However, due to the fact that the HILO factor is not equally important to all stocks and industry portfolios, a more detailed analysis of the coefficient estimates is needed in order to explain the suggested causality.

(36)

correlation (>0,8), I conclude that there is no reason to exclude or combine the variables due to the presence of multicollinearity.

Table 7. Correlation Matrix – Sweden & The Netherlands

!_!-!_! SMB HML HILO

!!-!! 1 ₁

SMB -0,1778 _0,0195 1 ₁

HML 0,1302 _0,2947 -0,0361 _-0,1045 1 ₁

HILO 0,4537 _0,6525 _-0,21960,0168 -0,0791 _0,4059 1 ₁

The upper numbers refer to the correlation results for the Swedish sample and the bottom numbers refer to the correlation results for the Dutch sample.

5. CONCLUSION

(37)

substantial amount of cases. However, the results are found to be stronger for industry sector returns than for individual stock returns.

Overall, the results suggest that an easy-to-calculate inflation-beta factor, HILO, can be used to extent current asset-pricing models, such as the original FF3FM, in order to provide a more accurate cost of equity estimation for individual stocks as well as industries. Moreover, the results suggest that investors should adjust their investment strategies and hedging strategies according to the desired inflation exposure risk level. However, due to the fact that the HILO factor is not equally important to all stocks and industry portfolios, a more detailed analysis of the coefficient estimates is needed in order to explain the suggested causality. Hence, I also leave open the question why company’s inflation exposures differ in the first place.

Limitations and future research

(38)

REFERENCES

Adler, M., & Duma, B. (1983). International portfolio choice and corporation finance: A synthesis. The Journal Of Finance, 38(3): 925-984.

Ang, A., Brière, M., & Signori, O. (2012). Inflation and individual equities.

Financial Analysts Journal, 68(4): 36-55.

Banz, R.W. (1981). The relationship between return and market value of common stocks. Journal of Financial Economics, 9(1): 3-18.

Bekaert, G., & Wang, X. (2010). Inflation risk and the inflation risk premium.

Economic Policy, 25(64): 755-806.

Bartholdy, J., & Peare, P. (2005). Estimation of expected return: CAPM vs. Fama and French. International Review Of Financial Analysis, 14(4): 407-427.

Bodie, Z. (1976). Common stocks as a hedge against inflation. The Journal Of

Finance, 31(2): 459-470.

Boons, M., De Roon, F. A., & Szymanowska, M. (November 7, 2011). Sorting out inflation. Working paper.

Boons, M., De Roon, F. A., & Szymanowska, M. (November 18, 2012). Sorting out inflation. Working paper.

Brounen, D., De Jong, A., & Koedijk, K. (2004). Corporate finance in europe: Confronting theory with practice. Financial Management (Blackwell

Publishing Limited), 33(4): 71-101.

Bruner, R. F., Eades, K. M., Harris, R. S., & Higgins, R. C. (1998). Best practices in estimating the cost of capital: Survey and synthesis. Financial Practice &

Education, 8(1): 13-28.

Carhart, M. (1997). On persistence in mutual fund performance. The Journal Of

Finance, 52(1): 57-82.

Chen, N.-F., Roll, R., & Ross, S. A. (1986). Economic forces and the stock market.

Journal Of Business, 59(3): 383-403.

Duarte, F.M. (December 10, 2010). Inflation risk and the cross section of stock returns. Working paper, MIT.

Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work. The Journal Of Finance, 25(2): 383-417.

Fama, E. F. (1981). Stock returns, real activity, inflation, and money. American

(39)

Fama, E. F. (1991). Efficient capital markets: II. The Journal Of Finance, 46(5): 1575-1617.

Fama, E. F., & French, K. R. (1992). The cross-section of expected stock returns.

The Journal Of Finance, 47: 427–465.

Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33(1): 3-56.

Fama, E. F., & French, K. R. (1996). The CAPM is wanted, dead or alive. The

Journal Of Finance, 51(5): 1947-1958.

Fama, E. F., & French, K. R. (1997). Industry costs of equity. Journal of Financial

Economics, 43(2): 153-193.

Fama, E. F., & MacBeth, J. D. (1973). Risk, return, and equilibrium: Empirical tests. Journal of Political Economy, 3:753–55.

Fama, E. F., & Schwert, O. (1977). Asset returns and inflation. Journal of Financial

Economics, 5(2): 115-146.

Firth, M. (1979). The relationship between stock market returns and rates of inflation. The Journal Of Finance, 34(3): 743-749.

Fisher, I. (1930). The theory of interest. New York: MacMillan.

Gallagher, L. A., & Taylor, M. P. (2002). The stock return-inflation puzzle revisited.

Economics Letters, 75(2): 147-156.

Geske, R., & Roll, R. (1983). The fiscal and monetary linkage between stock returns and inflation. The Journal Of Finance, 38(1): 1-33.

Graham, J. R., & Harvey, C. R. (2001). The theory and practice of corporate finance: Evidence from the field. Journal of Financial Economics, 60(2/3): 187-243. Groenewald, N., & Fraser, P. (2000). Forecasting Beta: How well does the

‘five-year rule of thumb’ do?. Journal of Business Finance & Accounting, 27: 953-982.

Hamberg, M. (2001). Strategic financial decisions. Malmo: Liber AB.

Hibbert, M., & Lawrence, E. R. (2010). Testing the performance of asset pricing models in different economic and interest rate regimes using individual stock returns. The International Journal of Banking and Finance, 7(1): 79-98. Hillier, D., Ross, S., Westerfield, R., Jaffe, J., & Jordan, B. (2010). Corporate

Finance – European Edition. Berkshire: McGraw-Hill Higher Education. Jaffe, J. F., & Mandelker, G. (1976). The "fisher effect" for risky assets: An empirical