Should we invest regionally?
Author: Michel Hietbrink Supervisor: Auke Plantinga
University of Groningen 4th of June 2020
Abstract:
In modern day portfolio theory, investing is all about diversification. Portfolios should be globally diversified to obtain the optimal risk/return trade off. However, people cannot connect with such a
portfolio. Familiarity breeds investment, therefore I have setup a regional portfolio including only companies operating in the 3 northern provinces of the Netherlands. The performance of the regional
portfolio is compared to the performance of the MSCI world index. I conclude that the regional portfolio has a positive pricing error. Moreover, the regional portfolio has a higher risk-adjusted return relative to the MSCI world index. This conclusion contradicts existing theories of diversification.
Introduction
In portfolio theory, most theories are based on the assumption that investors are rational. Moreover, the basis for most portfolio theories is the capital asset pricing model (CAPM) which additionally assumes investors to be risk averse as well. So investors are assumed to accept additional risk on their investments only if they are compensated for it by higher expected returns. Portfolio theories show that an investor can get a more beneficial risk/return trade-off by investing in a large portfolio of stocks. By ensuring that the portfolio consists of a diversified set of stocks, investors can optimize the risk/return trade off. Moreover, the stock-specific risk diminishes and the risk of the portfolio converges to the risk of the market. Therefore diversification is one of the main themes in asset allocation and stock selection.
A large amount of research has been conducted on how to optimize the diversification of a portfolio. The risk of the portfolio is reduced when the common factors among the selected stocks is minimalized. So picking stocks from different markets and sectors reduces the risk. Additionally selecting stocks from multiple countries reduces the risk too, since the fortunes of different countries do not align all the time. Thus classic research suggest that diversification is optimized if the portfolio is globalized, so that every diversification opportunity can be seized (Solnik, 1974). However, a significant amount of research suggests that most investors do have a certain degree of ”home bias”. Their investment strategy involves investing more in local or regional stocks relative to foreign stocks than suggested by research on optimal diversification. So do French and Poterba (1991) show that equity portfolios are mostly hold domestically. For Japan, the US and the UK the percentage of domestically held equity portfolio’s is larger than 80%. They argue that the lack of diversification is a result of investor choices, rather than institutional constraints. Heathcote and Perry (2007) report that over the period of 1990 until 2004 only approximately 25% of the total value of assets owned by U.S. residents are foreign assets. Kang & Stulz (1997) also observe this home bias, they state that “even though the barriers to international investments have fallen dramatically, foreign ownership of shares is still extremely limited and much smaller than one would expect in the absence of barriers to international investment” Nevertheless, regional investing is considered sub-optimal behavior, only performed by naïve investors. Therefore the question arises, why do investors have a “home bias”?
international company, but tracking the company in terms of non-financial operations is difficult, especially for a large group of international companies. For instance, if an investor wants to solely socially responsible investments, a well-diversified portfolio is more difficult to set up and track relative to an regional portfolio.
Furthermore one could argue that a regional portfolio still could be well-diversified. Since globalization is ever-increasing around the world, barriers between markets and countries are diminishing. A portfolio of companies located in a certain region of a country could still consist of companies that are operating internationally. Moreover, a regional company can be a subsidiary of a large multinational enterprise. therefore the risks of the companies are not solely based on the market risk of the region they are located in. This raises the question if it is still necessary to make an puzzle of international stocks to obtain a well-diversified portfolio. Can a regional portfolio offer roughly the same risk-adjusted return as a globally diversified portfolio over a prolonged period of time?
This question is key in the debate if regional investing should even be considered. If a regionally invested portfolio has a significantly worse risk-adjusted return relative to a well-diversified portfolio, one should never invest regionally. But if the difference in performance of the portfolio is roughly negligible, there are definitely motives to invest regionally.
Especially at the current time of the corona crisis, where countries have closed their borders and the entire society is in lock down. Politicians call the crisis one of the most disastrous times in the last century, excluding times of war. The impact on the economy is huge, an economic recession all around the world and many people lost their jobs. A lot of companies are balancing on the verge of bankruptcy. Governments and politicians request their inhabitants to support the local companies by shopping locally as frequently as possible. However, we could also support the local economy and companies by investing regionally.
invest in the familiar while often ignoring the principles of portfolio theory. A regional portfolio could solve the latter factor. By investing in the regional portfolio the people invest in companies who are active in their region. The company can operate directly in the region or through subsidiaries. So their investment indirectly benefits the region. With the investment they support local businesses thus, they invest in the “job of their neighbor”. The regional portfolio gives the people more of a feel by their investment. Next to investing for the return they generate, there is also a more behavioral motive since they support the region by doing so, which importance has been stretched in the current corona crisis.
So investing regionally definitely has beneficial aspects, however as mentioned before, it contradicts the classical theories of global diversification. So the question remains:
Should we invest regionally?
Therefore we should test if a regional portfolio can offer roughly the same risk-adjusted return as a well-diversified portfolio, so that the benefits of regional investing do not come at a significant loss in performance. To test this, I have set up a regional portfolio covering the region of the northern part of the Netherlands (e.g. the provinces Drenthe, Friesland and Groningen). The portfolio consist solely out of stock listed companies that are directly operating in the region, or indirectly through subsidiaries over the period of 1996 till 2017. The performance of the regional portfolio is compared to the MSCI world index. The MSCI world index is commonly used as a benchmark for global stock funds. intended to represent a broad cross-section of global markets.
Figure 1: The economics growth per province in the Netherlands, in % change in volume relative to last years value. (CBS, de regionale economie 2018)
Moreover, the northern region is characterized by relative low business confidence. However, the business confidence in the northern provinces is increasing relative to the other provinces. Furthermore, the unemployment rate in Groningen the highest out of all provinces in 2018, with a value of 5,1 % of the total working population in Groningen. However the unemployment rate is dropping significantly over the previous years with a drop of 3,4% since 2015.
inhabitants have the desire to split Friesland from the Netherlands, so it becomes its own country.
A critical note on this research is that I am not trying to prove any current portfolio theories to be wrong. Nor do I want to prove that regional investing is the best investment strategy available. What I am trying to prove is that regional investing can be a feasible option for individual investors. A regional portfolio could possibly help solve phenomenon’s in financial markets such as the limited participation puzzle. If the portfolio does not have a significantly different performance relative to a global portfolio, one should consider regional investing as an option. Moreover, the results of this research do not have hard implications, it is a first step in regional investing research. It shows that for a random area in the world, a portfolio with firms only operating in that area can also give you a quite well-diversified portfolio. This does not mean it will hold for any area in the world, but it could be a first step towards a change in current investing perceptions.
Previous Literature
To answer the proposed research question, we first should examine what makes a portfolio diversified. First of all, we should define the amount of stocks necessary for a well-diversified portfolio, Statman (1987) argued that for borrowing investors a well-well-diversified portfolio should include a minimum of 30 stocks, for lending investors the minimum is 40 stocks. According to Solnik (1974) the stocks selected for the well-diversified portfolio should be globally spread. He argues that the benefits of international diversification are so large, one should never invest regionally. The stock prices in 2 separate countries are barely correlated to one another, yielding more optimal diversification, however international investments does result in some exchange rate risk is .
Investors do not only expect higher returns, Ivkovich and Weisbenner (2003) show that they also realize it. According to their research the average household generates an additional return of 3,7% per year on their local stocks relative to its non-local stocks. They argue that in addition to the familiarity side of regional investing, information asymmetry also is a driver. They suggest that local investors are able to exploit local knowledge, especially on less known stocks which are not listed on the S&P 500. Coval and Moskowitz (2001) find similar results for mutual funds, where mutual fund managers earn abnormal high returns on nearby investments. They also argue that the phenomenon is caused by an informational advantage. Massa and Simonov (2006) find similar results for Swedish investors, suggesting that investors invest in stocks closely related to their non-financial income instead of hedging. Explaining the behavior with the familiarity motive. Moreover, they show that the familiarity based investing is information driven. The investments earn higher returns than when the investor would have hedged. Therefore we can conclude that there is already evidence that regional investing does not directly imply worse portfolio performance. Moreover, diversification can even destroy shareholder value, as is found by Doukas & Kan (2006) however it does not destroy the overall value of the company. On the other hand Seasholes & Zhu (2010) contradict the findings of the information advantage on local stocks. They show that purchases of local stocks significantly underperform sales of local stocks. And so found evidence that individual households do not have value-relevant information on local stocks, and so cannot obtain excess return on their investment by investing regionally.
Moreover, one could argue that over the decade information has become more easily available and attainable for investors all over the world and therefore the information advantage of local investors to diminish. Nevertheless the home bias is still present among investors. Therefore additional research on regional investing in the last decade can provide us more up to date knowledge.
Data & Methodology
limited data on business establishments. Since the regional portfolio requires precise information per business establishment, the ABR is not a well suited option (Lissen, 2004). The LISA register provides more in depth and precise statistics per business establishment relative to the HR of the VVK. Especially on the number of employees per establishment LISA provides the best information out of the registers. As LISA states on their website: “LISA serves an important role in the information requirement for social-economic and spatial economic research. No other source can provide you the same information”. Therefore the LISA register is used to analyze the business establishments in the region.
Figure 2: The headquarter location of the 168 stock listed companies operating in the Northern Netherlands.
Operating Sector
#Retail (excl. vehicles) 16
Manufactering of chemical products 14
Manufactering of foods 11
Wholesale and trade negotiators (excl. vehicles) 10
Architects, engineers and technical design and advice 8
Manufactering of other machinery and devices1 8
Manufactering of computers and electronical and optical devices 8
Service activities in information technology 7
Manufactering of other goods2 7
Manufactering of paper, cardboard and paper-and cardboard products 5
Manufactering of rubber-and plastic products 5
Table 1: The main operating sectors and number of companies per sector of the stock listed companies operating in the Northern Netherlands. (Operating Sector in which at least 5 firms are active in the sector)
1: Machinery and devices other than computer and technological and optical devices
2: Goods other than chemical products, food, paper and cardboard, rubber and plastic, tabaco, vehicles, metals in primary form, metal products, transport equipment, pharmaceutical materials and products, non-metallic mineral products.
These 168 companies will be used to create the regional portfolio. Two versions of the regional portfolio will be created, namely an equally weighted portfolio and a market capitalization weighted portfolio. The market capitalization weighted portfolio is a mainstream setup of a portfolio for large investors and investment firms. It follows the same weighting as all indexes around the world. The impact of an individual stock’s price change on the overall price change of the portfolio is proportional to the value of the market capitalization relative to market capitalization of the entire portfolio. The equally weighted portfolio is more mainstream for the small individual investor as do Ran and Levy (2009) show. They argue that investors tend to ignore the theoretical approaches as the Markowitz diversification in favor of the naïve diversification strategy called the Talmudic diversification. Moreover, they show that for small portfolios it outperforms the Markowitz diversification, but for large portfolios the Markowitz diversification is superior. For all the 168 companies included in the regional portfolios, I collect monthly total return data from January 1996 until November 2019. The total return includes interest, capital gains, dividends and distributions realized over the respective month. Additionally yearly market capitalization data was collected for the 168 companies for the market cap weighted portfolio. Yearly data since the rebalancing of the regional portfolio is done on a yearly basis. All the data is collected from the Refinitiv Eikon database. The market capitalization weighted portfolio is setup as the summation of all stock listed companies respective returns times their respective share in the total sum of the market capitalizations of all companies in the portfolio. The market capitalization weighted portfolio has more similar characteristics as a stock index relative to the equally weighted portfolio, since larger firms have a larger influence on the overall return of the portfolio, the same holds for a stock index. Not all 168 companies are operating in the region for the entire sample period, therefore the companies are only included in the portfolio if they are active in the region in that year. So the portfolio will be rebalanced on a yearly basis. This frequency is chosen because the LISA data in only available on a yearly basis. Resulting in a minimum amount of companies included in the regional portfolio of 44 in the year 1996, and a maximum amount of 144 in the year 2019. The equally weighted portfolio as well as the market capitalization portfolio weighted portfolio of 2017 will be held constant for 2 additional years to measure the performance in the most recent times.
different currency, the monthly total return will be adjusted with the change in the exchange rate of the respective currency with the euro of that specific month.
The performance of the regional portfolio has to be compared with the performance of a optimally well- diversified portfolio with no international barriers. This well-diversified portfolio is represented by the MSCI world index. The MSCI world index is commonly used as a benchmark for global stock funds intended to represent a broad cross-section of global markets. For the performance measure of the MSCI World index monthly total return data of the MSCI world index (expressed in dollars and adjusted with the change in exchange rate of the Dollar relative to the Euro for the respective month) is collected from the website of the MSCI for the period of January 1996 until November 2019. Additionally MSCI Europe, MSCI Pacific and MSCI North America total return data was acquired for this period. The 3 main regions data of the MSCI was acquired to obtain the dependence of the regional portfolios on the 3 main regions of the financial world.
To test the financial performance of the portfolios, we use the classical Capital Asset Pricing Model (CAPM) as well as the Fama & French 5-factor model (FF5). The Fama & French 5-Factor model is an extension of the CAPM, which is the common basic theory used to measure the financial performance of a portfolio. However the CAPM is not able to explain portfolio returns completely, resulting in significant pricing errors. To reduce these pricing errors of the CAPM, Fama & French set up a model with 4 additional factors which help to explain diversified portfolio returns. The model has the following form:
𝑟𝑝− 𝑟𝑓= 𝛼𝑖+ 𝛽1(𝑟𝑚− 𝑟𝑓) + 𝛽2 𝑆𝑀𝐵 + 𝛽3 𝐻𝑀𝐿 + 𝛽4 𝑅𝑀𝑊 + 𝛽5 𝐶𝑀𝐴 + 𝑒𝑖 (1) Where 𝑟𝑝− 𝑟𝑓= 𝛼𝑖 + 𝛽1(𝑟𝑚− 𝑟𝑓) is the Capital Asset Pricing Model.
The rp is the return of the portfolio, rf is the risk-free rate, rm – rf is the market risk premium.
SMB is the return of a diversified portfolio of small stocks minus the return of a diversified portfolio of big stocks, i.e. the “size effect”. HML is the return spread on diversified portfolios of high book-to-market value and low book-to market value stocks, i.e. the “value effect”. RMW is the return spread on diversified portfolios with robust and weak profitability. The CMA is the return spread on diversified portfolios of stocks who invest conservatively minus stocks who invest aggressively. αi captures the remaining pricing error of the FF5, and ei is the
The monthly data for all the factors of the model are acquired from the Fama & French developed 5 factors dataset on the Fama & French website. The developed dataset is used since this dataset represents the global value’s of the factors. This suits best for a globally diversified portfolio and therefore suits the MSCI world index. Moreover, it also suits the regional portfolio, since the portfolio includes companies from all over the world as can be seen from figure 2. The data on the Fama & French website is expressed in dollar therefore, the data is adjusted with the respective change in the dollar/euro exchange rate per month. For a well-diversified model, the FF5 will capture most of the pattern in the returns and will so have a low pricing error and thus, a small “α” coefficient. If a portfolio has a positive “α” coefficient, the portfolio outperforms the market, and the better performance cannot be explained by any of the 4 additional factors. For a negative “α” it is vice versa. Therefore the alpha is a suitable measure for the overall performance of the portfolio. However, the portfolio could still have additional risk for any “α” coefficient. Therefore, the risk-adjusted returns of the regional portfolio should be compared to the well-diversified portfolio.
A suitable measure to compare the risk-adjusted returns of the portfolio’s is the Sharpe Ratio. It is one of the most referenced risk adjusted return measures used in finance. To calculate the Sharpe ratio, the following formula is used:
𝑆𝑝 = 𝑟𝑝−𝑟𝑓
𝜎𝑟𝑝 (2)
Where Sp is the Sharpe ratio coefficient, rp is the mean return of the portfolio over the sample
period, rf is the mean risk-free rate over the sample period. σrp is the standard deviation of the
return of the portfolio over the sample period. The Sharpe ratio gives a coefficient which resembles the return per “unit of risk”. So it is a return adjusted for risk. Therefore it makes an easy comparison between the performance of different portfolios.
The one assumption necessary for the Sharpe ratio to be a valid measure is the assumption that the returns of the respective portfolio are normally distributed. This is usually the case when using logarithmic and so continuously compounded returns. Therefore the returns of the equally weighted portfolio, the market cap weighted portfolio and the MSCI world index will all be computed in to logarithmic returns. The logarithmic returns will be used in the regression of the Fama French 5-factor model as well as in the calculation of the Sharpe ratio. To test if the Sharpe ratio per portfolio are statistically different from each other, I use the Jobson & Korkie test (Jobson & Korkie, 1981). The test statistic has the following formula:
With 𝛳 = (1 𝑇) [ 1 2 (𝑆𝐴 2+ 𝑆 𝐵2− 𝑆𝑎𝑆𝑏(1 + 𝜌𝐴𝐵2 )) + 2(1 − 𝜌𝐴𝐵)]
Where SA and SB are the Sharpe ratio’s of portfolios A and B, and ρAB is the correlation
coefficient between returns of portfolio A and B. T is the number of return intervals. The test statistic Z is approximately normally distributed with a mean of zero and a standard deviation of one for large sample sizes.
Since the Sharpe ratio resembles a risk-adjusted return and diversification minimizes the risk of a portfolio. One would expect the most well-diversified portfolio have the highest Sharpe ratio. So the MSCI world index which resembles the optimally diversified portfolio, should have the highest Sharpe ratio relative to the two regional portfolios. Moreover, one would expect the MSCI world index to not have any pricing error, since the FF5 is setup to accurately predict the returns of such a well-diversified index. For the two regional portfolios, especially the market capitalization weighted portfolio, we expect quite similar findings in term of FF5 coefficients, since the two portfolios both should be roughly as diversified as the MSCI. For the equally weighted portfolio the coefficients could slightly differ since it has relatively large weights for relatively small firms, therefore the portfolio returns could be relying relatively more on the factors such as SMB, which are included in the FF5.
Table 2 summarizes all the data used in the Fama French 5-factor model as well as for the Sharpe ratio, so the data of the regional portfolio’s returns, the MSCI world returns, the Market risk premium, the risk-free rate and the SMB, HML, RMW and CMA factor.
Variable Obs Mean Std. Dev. Min Max
Regional Return 287 1.20% 4.65% -21.69% 18.42%
Regional Return (MC) 287 0.77% 4.05% -15.37% 10.27%
MSCI World return 287 0.63% 4.29% -18.74% 10.70%
Market risk premium 287 0.56% 4.29% -17.61% 11.39%
Risk-free rate 287 0.18% 0.17% 0.00% 0.54%
SMB 287 0.07% 1.96% -8.60% 8.32%
HML 287 0.26% 2.47% -10.03% 12.53%
RMW 287 0.38% 1.49% -5.71% 6.34%
CMA 287 0.20% 1.94% -6.44% 9.81%
MSCI Europe return 287 0.63% 5.03% -21.29% 13.24%
MSCI Pacific return 287 0.29% 4.81% -17.61% 16.10%
MSCI North America return 287 0.76% 4.33% -17.64% 10.73%
By having a glance at the statistics in table 2, we find quite interesting results. Firstly the equally weighted regional portfolio has the highest mean return by quite a significant margin. However, the portfolio also has the highest standard deviation of the 3 portfolios, which is in line with theory on being rewarded for bearing extra systematic risk. The market capitalization weighted regional portfolio has a return quite similar as the return of the MSCI world index, but it is still slightly higher relative to the MSCI world index. However we cannot jump into conclusions solely based on these findings.
Results
First we test our data for autocorrelation and heteroskedasticity. For all three portfolio’s we find no evidence of autocorrelation but do find heteroskedasticity among all three (Appendix). Therefore in all the CAPM and Fama French 5-factor regressions we use heteroskedasticity-consistent standard errors.
Capital Asset Pricing Model
First I run the CAPM regression on the logarithmic returns of the equally weighted regional portfolio, the market capitalization weighted portfolio and the MSCI world index. Table 3 provides the results of the regression.
(1) (2) (3)
VARIABLES EWregional MCWregional MSCIworld
Beta 0.939*** 0.808*** 0.996*** (0.043) (0.037) (0.009) Alpha 0.005*** 0.001 -0.001*** (0.001) (0.001) (0.000) Observations 287 287 287 Adjusted R-squared 0.751 0.734 0.987
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table 3: Results CAPM regression on the 3 portfolios, where EWregional is the equally weighted regional portfolio, MCWregional is the market capitalization weighted portfolio and MSCIworld is the MSCI world index.
portfolio and the MSCI world index the coefficient is statistically different from zero at the 1% significance level. Thus we can only conclude that the equally weighted portfolio is outperforming the market and the MSCI is ever so slightly underperform the market according to the CAPM. For the market capitalization weighted regional portfolio the coefficient is not statistically different from zero, therefore we cannot conclude that the performance of the portfolio is different from the performance of the market. What we are more interest in is the performance of the regional portfolios relative to the MSCI world index.
We test if the alpha of the portfolios are different from each other with the Student’s T-test. We find that the alpha of the equally weighted portfolio is statistically different from the alpha of the MSCI of the at the 1% significance level, for the market capitalization weighted regional portfolio it is statistically different from the alpha of the MSCI world index at the 10% significance level (appendix). Therefore we can conclude that the both regional portfolios are outperforming the MSCI world index in returns terms.
Furthermore can be seen that the Market Beta of the MSCI world index roughly equals 1, indicating that the returns of the MSCI world index move one for one with the market risk premium. This is to be expected since the Fama & French data for the market risk premium of developed countries is most likely depending on roughly the same underlying assets as the MSCI world index. This also explains the high adjusted R2 found for the MSCI world index
Fama & French 5-factor model
Table 4 presents the results of the Fama & French 5-factor model for all 3 portfolios.
(1) (2) (3)
VARIABLES EWregional MCWregional MSCIworld
Beta 0.959*** 0.833*** 1.003*** (0.041) (0.037) (0.006) SMB 0.150** -0.217*** -0.181*** (0.072) (0.065) (0.010) HML 0.263*** 0.043 0.022* (0.085) (0.077) (0.012) RMW 0.171 0.126 0.040*** (0.107) (0.097) (0.015) CMA -0.176 0.019 0.011 (0.119) (0.107) (0.016) Alpha 0.004** 0.001 -0.001*** (0.001) (0.001) (0.000) Observations 287 287 287 Adjusted R-squared 0.764 0.747 0.995
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table 4: Results FF5 regression on the 3 portfolios, where EWregional is the equally weighted regional portfolio, MCWregional is the market capitalization weighted portfolio and
MSCIworld is the MSCI world index.
First of all, from table 4 can be seen that for every portfolio the adjusted R-squared is higher relative to the CAPM regression results, so the FF5 is indeed better at explaining the variance in the return of a stock or portfolio. Similarly as in the CAPM regression for the equally weighted regional portfolio the alpha is positive and the coefficient is statistically different from zero at the 5% significance level. The market cap weighted regional portfolio has an alpha that is not statistically different from zero at any significance level. The MSCI world has a negative alpha, the coefficient is statistically different from zero at the 1% significance level. So the FF5 yields the same conclusions of the CAPM, but with a higher precision. However, only the alpha of the equally weighted portfolio is statistically significantly different from the alpha of the MSCI world index. So the equally weighted portfolio is also according to the FF5 model outperforming the MSCI world index. For the market cap weighted portfolio the returns are not significantly different from the returns of the MSCI world index.
As can be seen from the additional 4 factors included in the FF5 regression, the equally weighted portfolio has indeed relatively large coefficient for the 4 factors. However, most of the coefficient are not statistically different from zero. The dependence on the 4 factors decreases when the portfolio is market cap weighted, as is expected.
For the market cap weighted regional portfolio as well as the MSCI world index the alpha of the CAPM as well of the FF5 are very small, roughly negligible in economic significance. However the alpha of the equally weighted portfolio is quite significantly large in both the CAPM and FF5 regression. An explanation for this could be the data of the market risk premium as well as the 4 additional factors acquired from the Fama & French website. The data resembles the developed countries data for these factors. The equally weighted portfolio could be quite skewed towards the European region through the equally weighting of all stocks. Therefore the acquired factor data could be imprecise in explaining the respective return of the portfolio, resulting in a pricing error. To test for this we regress the excess return for all three portfolios against the excess return of the MSCI Europe, MSCI Pacific and the MSCI North America. Table 4 presents the results of the three regressions.
(1) (2) (3)
VARIABLES EWregional MCWregional MSCIworld
MSCIeu 0.428*** 0.280*** 0.298*** (0.052) (0.044) (0.004) MSCIpac 0.028 0.002 0.153*** (0.041) (0.035) (0.003) MSCIna 0.455*** 0.527*** 0.548*** (0.059) (0.050) (0.005) Constant 0.006*** 0.002 -0.000 (0.001) (0.001) (0.000) Observations 287 287 287 Adjusted R-squared 0.761 0.770 0.998
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table 4: The world 3 main regions dependence per portfolio over the sample period of 1996 until 2019, where EWregional is the equally weighted regional portfolio, MCWregional is the market capitalization weighted portfolio and MSCI is the MSCI world index.
world best, as could be seen from the adjusted R2 of the MSCI world regression for the CAPM
and FF5. Sharpe ratio
The CAPM and the Fama & French 5-factor model only give a measure on the return performance of a portfolio. It shows us if the returns can be contributed to certain pattern in stock price returns or that the return on a stock or portfolio really outperforms the market. However the return performance is not adjusted for risk, so if a portfolio outperformed the market, it could still be more risky than the market and therefore, not be the preferred investment. The Sharpe ratio gives a good measure of the risk-adjusted return a portfolio and is easily comparable between portfolio. Table 5 contains the results of the Sharpe ratio per portfolio over the time period of 1996 until 2019.
Variable EWregional MCWregional MSCI world
Sharpe Ratio 0.258 0.190 0.147
Table 5: The Sharpe ratio per portfolio over the sample period of 1996 until 2019, where EWregional is the equally weighted regional portfolio, MCWregional is the market capitalization weighted portfolio and MSCI is the MSCI world index.
world. Therefore, the resulting “regional” portfolio is also a well-diversified portfolio. This minimizes the differences between the regional portfolios and the MSCI world index in risk terms. Nevertheless, the significantly lower standard deviation relative to the MSCI is a striking result, which lacks a clear explanation.
The Jobson & Korkie test is used to test if the Sharpe ratios are statistically significantly different from each other. The Z-statistic from the Jobson & Korkie test is 223.19 for the difference between the equally weighted regional portfolio and the MSCI world index Sharp ratios. For the difference in Sharpe ratios of the market capitalization weighted regional portfolio and the MSCI world index the Z-statistic is 47.79. Both Z-statistic correspond to a probability of 0.00000 of the Sharpe ratios being equal. Therefore we can conclude that the Sharpe ratios of the regional portfolios are both statistically significantly different from the Sharpe ratio of the MSCI world index at the 1% significance level. So we can conclude that the Sharpe ratios of both regional portfolios are indeed higher than the Sharpe ratio of the MSCI world index.
Main Findings
The Capital Asset Pricing Model and the Fama & French 5-Factor model regression both show a statistically significant positive pricing error for the equally weighted regional portfolio. Thus the equally weighted regional portfolio is outperforming the market. For the market capitalization weighted portfolio I do not find an alpha significantly different from zero, therefore I cannot conclude that it is underperforming nor outperforming the market. The alpha of the MSCI world index is significantly different from zero and has a negative value, however the value is too small to have any economic significance.
The most striking finding is the Sharpe ratio of the 2 regional portfolios relative to the MSCI world index. The 2 regional portfolios have a dramatically larger Sharpe ratio relative to the MSCI, indicating that the risk-adjusted return is higher for the 2 regional portfolios relative to the risk-adjusted return of the MSCI. Especially since the higher Sharpe ratios for the market capitalization regional portfolio is partially due to a lower standard deviations relative to the MSCI world index. Therefore it could be argued that the portfolio is bears less risk relative to the MSCI world index. This is in contrast with the main theory of Solnik’s global diversification in portfolio theory. These findings are more difficult to explain however, I will provide multiple possible explanations for these findings.
Conclusion
The adjusted returns for the regional portfolios is significantly higher than the risk-adjusted return of the MSCI world index. Therefore we can conclude that the performance in financial terms of the regional portfolios is at least as good as that of a well-diversified portfolio. This contradicts classic well-known theories on diversification.
Only the equally weighted regional portfolio has a statistically significant positive alpha in both the CAPM model as the FF5-factor model. However, we cannot directly conclude that the portfolio outperforming the market, since the pricing error could also be caused by the data on the factors of the CAPM and FF5 model. The market capitalization weighted regional portfolio does not have an alpha significantly different from zero for both models. The alpha of the MSCI world index is statistically significant and negative in both models, however the coefficient is so small that the economic significance is negligible.
The Sharpe ratios show that a regional portfolio can still be a well-diversified portfolio, even though the regional portfolio contains solely companies which are directly operating in the norther Netherlands or indirectly through subsidiaries. As can be seen from the companies included in the regional portfolio, which headquarters are scattered all over the world. The one downside to regional investing as discussed beforehand was the potential loss in financial performance. Investing regionally would imply investing in a less diversified portfolio, resulting in a riskier portfolios and therefore the risk-adjusted return of the portfolio would be lower than that of a well-diversified portfolio. As stated before, this assumption does not hold for the regional portfolios constructed in this research. So investing in the northern Netherlands regional portfolio has no downside relative to other investment opportunities. Thus, regarding the potential non-financial upsides of the regional investment, investment opportunities like the norther Netherlands regional portfolio should be made available as an option for everyone in the region or Netherlands. Inhabitants of the region will feel more connected to the their investment as they prefer to be, as stated by literature on familiarity and investment. Therefore the regional portfolio could attract people in the region to participate in the stock market and so partially solve the non-participation problem in the stock market. By investing regionally the people in the region indirectly support the region itself as well as the people working in the region. So you invest in the job of your neighbor. Especially in current times of the corona crisis the importance of people to support their region and the local companies is emphasized.
well-diversified portfolio, but does mean that investing in the company does not directly flow into the companies in the region. One could argue that investing in the regional portfolio will not result into a significant cash inflow into the companies and business establishments located in the region. However, if the investment into the regional portfolio has no downside in financial performance, why should you not invest in the portfolio if you are from the region? Even if the cash flow to the companies and business establishments in the region is limited, it cannot harm the companies and business establishments in the region. If you do not shoot, you will never score.
However, one should take in mind that this research is just one piece in the puzzle of regional investing. That these results hold for this region, does not imply that it will hold for any region in the world. Moreover, as stated before, the results do not imply that the classical theories on diversification are wrong. Diversification is still the most important aspect of portfolio theory, this research only shows that it is not necessary to make a difficult puzzle of companies operating around the world to optimally diversify your portfolio. This research does not try to indicate that everyone should invest regionally and that current investment strategies are less beneficial. The point made is that regional investing is not as bad as argued is certain literature. The research shows that investing in a regional portfolio can be a feasible option for investors and therefore it should be considered as an option among stock markets and their participants.
Direction for further research
As argued in the conclusion, this research is just one piece in the puzzle of regional investing. The research on regional investing should be expanded to additional regions and countries, that these findings hold for this region, it is not a given for other regions.
Other interesting setups would be equivalent portfolios as the on in this paper, but then with e.g. only companies with a certain minimum ESG score. In such a manner you support the companies and business establishments in the region and you support a the local economy to transition into a “greener” economy.
References:
Coval, Joshua D., and Tobias J. Moskowitz, 2001, The geography of investment: Informed trading and asset prices, Journal of Political Economy 109(4), 811-841.
Dimmocka S.G., R. Kouwenberg, , O. S. Mitchell, K. Peijnenburge, ‘Ambiguity aversion and household portfolio choice puzzles: Empirical evidence’, Journal of Financial Economics Volume 119, Issue 3, March 2016, Pages 559-577
Doukas, John A., and Ozgur B. Kan, 2006, ‘Does global diversification destroy firm value?’, Journal of International Business Studies, Vol. 37, pp. 352-371.
Fama, E. F., and French, K.R., 2015, “A five-factor asset pricing model”, Journal of Financial Economics, Volume 116, Issue 1, April 2015, Pages 1-22
French, K.R., and J.M. Poterba, 1991, ‘Investor diversification and international equity markets’, American Economic Review, Vol. 81, No. 2, pp. 222-226.
Heathcote and Perri, 2013, ‘The international diversification puzzle is not as bad as you think’, Journal of Political Economy, Vol. 121, No. 6, pp.
Het Centraal Bureau voor de Statistiek, 2019 “De regionale economie 2018”
Het LISA, VVK Handelsregister en CBS Bedrijvenregister met elkaar vergeleken : Vestigingen en banen naar economische activiteit. / Wissen, L. van. 2004.
Huberman, G., 2001, ‘Familiarity breeds investment’, Review of Financial Studies, Vol. 14, No. 3, pp. 659-680.
Ivkovic, Zoran, and Scott Weisbenner, 2005, Local does as local is: Information content of the geography of individual investors’ common stock investments, Journal of Finance 60, 267– 306.
Jobson, J. D., and Bob Korkie. ‘‘Performance Hypothesis Testing with the Sharpe and Treynor Measures.’’ Journal of Finance 36, no. 4 (September 1981a): 889–908.
Kang, J., and Stulz, R.,1997, “Why is there a home bias? An analysis of foreign portfolio equity ownership in Japan” Journal of Financial Economics, 46, 3-28.
Oehler, A., Rummer, M. Walker, T. and Wendt, S. 2010 ”Are Investors Home Biased? Evidence from Germany” DIVERSIFICATION AND PORTFOLIA MANAGEMENT OF MUTUAL FUNDS, Gregoriou, G. N., ed., Palgrave MacMillan, Houndmills, pp. 57-77, 2007
Seasholes, Mark S., and Ning Zhu, 2010, ‘Individual investors and local bias’, Journal of Finance, Vol. 65, No. 5, pp. 1987-2009.
Solnik, Bruno H., 1974, ‘Why Not Diversify Internationally Rather Than Domestically?’, Financial Analysts Journal, Vol. 30, No.4, pp. 48-52.
Appendix
Equally weighted regional portfolio:
Durbin-Watson d-statistic( 6, 287) = 1.836645
White's test for Ho: homoskedasticity against Ha: unrestricted heteroskedasticity
chi2(20) = 66.04 Prob > chi2 = 0.0000
Market Cap weighted regional portfolio:
Durbin-Watson d-statistic( 6, 287) = 2.145716
White's test for Ho: homoskedasticity
against Ha: unrestricted heteroskedasticity
chi2(20) = 75.29 Prob > chi2 = 0.0000
MSCI world index:
Durbin-Watson d-statistic( 6, 287) = 2.402797
White's test for Ho: homoskedasticity against Ha: unrestricted heteroskedasticity
chi2(20) = 118.94 Prob > chi2 = 0.0000
MCW regional
MSCI world
Test for CAPM if alpha of equally weighted regional portfolio is different form alpha MSCI: ( 1) _cons = -.0011056
F( 1, 285) = 18.21 Prob > F = 0.0000
Test for CAPM if alpha of MC weighted regional portfolio is different form alpha MSCI: ( 1) _cons = -.0011056 F( 1, 285) = 3.83 Prob > F = 0.0512 ` _cons .0014231 .0012916 1.10 0.271 -.0011191 .0039654 MktRF .8083524 .0370225 21.83 0.000 .7354801 .8812246 LRegRRFmc Coef. Std. Err. t P>|t| [95% Conf. Interval] Robust Root MSE = .0209 R-squared = 0.7345 Prob > F = 0.0000 F(1, 285) = 476.73 Linear regression Number of obs = 287
Fama & French 5-factor model: EWregional: MCWregional: MSCI World: _cons .0037592 .0014585 2.58 0.010 .0008882 .0066302 CMA -.1764202 .1188444 -1.48 0.139 -.4103585 .057518 RMW .171204 .1074572 1.59 0.112 -.0403194 .3827273 HML .2627609 .085226 3.08 0.002 .0949983 .4305234 SMB .1500984 .0719944 2.08 0.038 .0083815 .2918152 MktRF .9586329 .0406864 23.56 0.000 .8785442 1.038722 LRegRRF Coef. Std. Err. t P>|t| [95% Conf. Interval] Total .618032413 286 .002160952 Root MSE = .02257 Adj R-squared = 0.7643 Residual .14312348 281 .000509336 R-squared = 0.7684 Model .474908933 5 .094981787 Prob > F = 0.0000 F(5, 281) = 186.48 Source SS df MS Number of obs = 287
_cons .0008085 .0013165 0.61 0.540 -.001783 .0034 CMA .0189952 .1072754 0.18 0.860 -.1921702 .2301606 RMW .1256305 .0969967 1.30 0.196 -.065302 .316563 HML .0429746 .0769297 0.56 0.577 -.108457 .1944062 SMB -.2169245 .0649861 -3.34 0.001 -.3448458 -.0890032 MktRF .8329339 .0367257 22.68 0.000 .7606415 .9052263 LRegRRFmc Coef. Std. Err. t P>|t| [95% Conf. Interval] Total .468860571 286 .001639373 Root MSE = .02037 Adj R-squared = 0.7469 Residual .116614852 281 .000414999 R-squared = 0.7513 Model .352245719 5 .070449144 Prob > F = 0.0000 F(5, 281) = 169.76 Source SS df MS Number of obs = 287
Test for FF5 if alpha of equally weighted regional portfolio is different form alpha MSCI: ( 1) _cons = -.0012554
F( 1, 281) = 11.82 Prob > F = 0.0007
Test for FF5 if alpha of MC weighted regional portfolio is different form alpha MSCI: ( 1) _cons = -.0012554
F( 1, 281) = 2.46 Prob > F = 0.1181
Region Dependence regression: EW regional: MCW regional: _cons .0057163 .0013586 4.21 0.000 .0030421 .0083904 LMSCInaRF .4554407 .0587366 7.75 0.000 .3398247 .5710567 LMSCIpacRF .0284735 .0408248 0.70 0.486 -.0518852 .1088322 LMSCIeuRF .4278376 .051693 8.28 0.000 .3260861 .5295891 LRegRRF Coef. Std. Err. t P>|t| [95% Conf. Interval] Total .618032413 286 .002160952 Root MSE = .02271 Adj R-squared = 0.7613 Residual .145981311 283 .000515835 R-squared = 0.7638 Model .472051102 3 .157350367 Prob > F = 0.0000 F(3, 283) = 305.04 Source SS df MS Number of obs = 287
MSCI world:
Correlation between returns:
_cons -.0001191 .0001052 -1.13 0.258 -.0003261 .0000879 LMSCInaRF .5476867 .0045469 120.45 0.000 .5387366 .5566367 LMSCIpacRF .1533336 .0031603 48.52 0.000 .1471129 .1595543 LMSCIeuRF .2981814 .0040016 74.51 0.000 .2903047 .3060582 LMSCIRRF Coef. Std. Err. t P>|t| [95% Conf. Interval] Total .529183352 286 .001850291 Root MSE = .00176 Adj R-squared = 0.9983 Residual .000874803 283 3.0912e-06 R-squared = 0.9983 Model .528308549 3 .17610285 Prob > F = 0.0000 F(3, 283) = 56969.50 Source SS df MS Number of obs = 287
LMSCIRRF 0.8681 0.8730 1.0000 LRegRRFmc 0.9044 1.0000
LRegRRF 1.0000
