Should You Tilt Your Equity Portfolio to Smaller Countries?

Abstract Should You Tilt Your Equity Portfolio to Smaller Countries? This paper examines the relation between country size, measured as the aggregate market capitalization of the listed stocks in a country, and individual stock returns. We find that stocks from small countries tend to have higher average returns than stocks from large countries. The country size effect is largely independent of the firm size effect and other country quantitative factors such as book-to-market and momentum. We conjecture that the country-size effect is due to home bias and provide mixed evidence in support of this conjecture. Contact Info: Gregg S. Fisher Gerstein Fisher Head of Quantitative Research and Portfolio Strategy 565 Fifth Avenue, 27th Floor New York, NY 10017 Gfisher@gersteinfisher.com Ronnie Shah University of Texas at Austin - Department of Finance McCombs School of Business Austin, TX 78712 Ronnie.Shah@mccombs.utexas.edu Sheridan Titman 1 University of Texas at Austin - Department of Finance McCombs School of Business Walter W. McAllister Centennial Chair in Financial Services Austin, TX 78712 Sheridan.Titman@mccombs.utexas.edu Keywords: Size Premium, International Investing, Portfolio Construction 1 Contact Author. Sheridan Titman is an advisor to Gerstein Fisher, which employs quantitative equity investment strategies that tilt towards smaller countries. The views expressed here are those of the authors and not necessarily those of any affiliated institution. We thank Chris Meeske, Ashvin Viswanathan, Tianyu Wang and other members of the Gerstein Fisher Investment Strategy & Research Group for their research assistance. This research has benefited from discussions with Rawi E. Abdelal of Harvard Business School. 1

Introduction US investors continue to shift investments into foreign markets. Indeed, over the past 10 years, US equity mutual funds have experienced $834 billion in net outflows, compared to $643 billion of net inflows into international funds. 2 These flows have increased the share of equity mutual funds assets that invest in international markets from 23% to 27%, and the average asset allocation fund currently invests 30% of all equity assets internationally. 3 The proliferation of longonly equity strategies that invest in multiple countries necessitates a country allocation policy that takes into account the trade-offs of investing in different financial markets. Traditional portfolio theories, like the CAPM, suggest that the value-weighted combination of the country portfolios provides an efficient allocation. However, a variety of frictions associated with international investing can lead to substantial deviations from value-weighting. Indeed, a number of authors starting with Keppler and Traub (1993, 2011) have argued that returns in smaller countries are higher than returns in larger countries, so tilting towards smaller country stocks may increase returns as well as provide diversification benefits. Other authors have observed a momentum effect in country indices, suggesting that one might not want to hold country weights constant. In this paper we will reexamine what we will be referring to as the small country effect, or the tendency for stocks in smaller financial markets to out-perform stocks in larger financial markets. To what extent is this driven by the inclusion of Japan, the largest foreign market, which also had terrible returns in the 1990s? To what extent is it driven by the fact that the average market capitalization of stocks in smaller countries tends to be smaller? Perhaps the small country effect is simply a manifestation of the small firm effect. Finally, we consider the relation between the country momentum effect and the small country effect. We also consider a potential explanation for the small country effect. Intuitively, if investors are risk averse and subject to home bias, the expected rates of return of stocks should be higher in smaller countries because investors in these countries, who have less diversification 2 Source: 2015 Investment Company Institute Factbook. 3 Source: Morningstar. Calculated using Total Net Assets from US and International Equity Mutual funds from December 2005 to December 2016. Asset Allocation data based on weighted average of US and non-us equity share of assets for mutual funds categorized by Morningstar as Asset Allocation/Target Date. 2

opportunities, require higher rates of return. Small country stocks may also attract fewer foreign investors, because the small countries receive less attention from sell-side analysts and they tend to be less regulated and provide lower investor protection. The smaller markets are also potentially more vulnerable to the risks associated with the fickle nature of global portfolio flows. Each of these effects is likely to be less important in developed markets that are more open to foreign investors, and the larger stocks in the smaller markets that have access to international investors are likely to be less affected. Hence, the small country effect should be strongest for the smaller stocks in the emerging markets, and increased globalization of financial markets should weaken the country size premium over time. To address these issues we analyze individual stock returns and measure the extent to which the size of the market in which the firm is domiciled influences returns after controlling for firm size. We find that there is a small country effect, and that it is not a manifestation of the small firm effect. These results are influenced by the presence of Japan, a large country with poor returns, but the results are still significant in a sample that does not include Japan. We also find that our results are not driven by other country-level factors such as momentum or the average book-to-market ratio in the country. Our evidence provides mixed support for our conjecture that the small country effect is generated because of home bias and market frictions that reduce the access of small country stocks to international investors. We find that the small country effect is less pronounced in the more recent period, which is consistent with the idea that impediments to international investing has been reduced over time. We also find that the small country effect is stronger in emerging markets, which is consistent with both international investors having limited access to these markets, as well as the fact that investors in these markets tend to have less ability to invest internationally. Our most puzzling result, however, is that the small country effect is as strong for large capitalization stocks as small capitalization and mid-capitalization stocks. However, at least part of this observation can be attributed to differences in analyst following, which may be a proxy for the interest in these stocks by international investors. Among large stocks, we find that stocks from smaller markets tend to have significantly lower analyst coverage when compared to stocks from larger markets. We do not find this relationship for small capitalization and mid-capitalization stocks, but the coverage of these stocks is fairly minimal. 3

In addition to the earlier cited papers that directly focus on the small country effect, the issues raised in this paper relate to the more general international investment literature. Keppler and Encinosa (1993, 2011) analyze 18 equity markets that are components of the MSCI World Index and show that a capitalization-weighted portfolio that invests in the six smallest markets has a 12.79% annualized compound return, outperforming the capitalization-weighted portfolio that invests in all 18 markets by 5.02% over the period January 1970 to December 2009. Asness, Liew and Stevens (1997) show that country-level size, momentum and value (aggregate price-to-book) explain differences in country returns. Desrosiers, L Her and Plante (2004) analyze the performance of global investment strategies based on country indices from the 18 largest stock markets and find that country momentum explains differences in country returns, but aggregate book-to-market does not. Li and Pritamani (2015) find country size and momentum effects for various emerging and frontier markets. Angelidis and Tessaromatis (2016) explore how to best construct a multi-country global portfolio using value, momentum, low volatility and size countrylevel variables. Our paper contributes to this literature in a number of ways, in addition to providing an explanation for the country size effect. First, larger countries tend to have larger firms - our regression methodology allows us to test whether the firm size effect drives the country size effect, while the majority of the previous literature in this area reports portfolio return differences for combinations of country indices. Second, we show that our results are not driven by Japan, which is the largest country but happens to have among the poorest returns in our sample. Third, we find evidence of a significant country size effect in both developed and emerging markets and show that this relationship has weakened over time. Last, we control for other country-level quantitative factors such as momentum and the book-to-market ratio, and our results generally suggest the country size premium is independent of these other sources of expected return. Our paper is also related to research on the benefits of international diversification. Asness, Israelov and Liew (2011) suggest that investors benefit from international diversification in the long-run as economic growth drives variation in country returns. Braymen and Johnson (2015) show that a trade-adjusted weighting scheme can be used to improve risk-adjusted performance relative to a GDP-weighted portfolio. Goetzmann, Li and Rouwenhurst (2005), Eun and Lee (2010), and Christoffersen, Errunza, Jacobs and Langlois (2011) suggest that emerging markets may be less integrated than developed markets and thus provide greater diversification benefits when compared to portfolios that consist of only developed market equities. Our paper, in contrast, 4

suggests that forming more diverse portfolios by under-weighting large countries and overweighting small countries not only improves diversification but can also increase expected returns. The rest of the paper is organized as follows: The first section explains the data sources used in this study and presents descriptive statistics of value-weighted portfolios of country indices. The second section provides country-level analysis on the country size premium. The third section discusses whether home bias explains the country size effect. The fourth section reviews evidence on analyst coverage and country size. The final section concludes. I. Data Sources and Summary Statistics Our research examines stocks from markets that MSCI either classifies as developed or emerging. 4 While the definitions are dynamic, we use the initial country classification at the beginning of the sample period, January 1990, to ensure no forward-looking bias. Our analysis also excludes various emerging market countries. We exclude Argentinian stocks due to the transition from a floating rate currency (prior to 1992) to a fixed rate currency (until 2001) and back to a floating rate currency. We exclude Chinese stocks, since until recently foreigners could not purchase these stocks due to government restrictions, and Canadian stocks, as a high proportion of those stocks are also traded on US exchanges. We also exclude stocks from Russia, the Czech Republic, Egypt, Qatar, United Arab Emirates and Colombia, which lack sufficient data, and we exclude ADRs, GDRs, and stocks that are headquartered in a country that is different from country of the stock exchange that the stock is listed on. Our final sample consists of 37 different international markets. For developed markets, our sample starts in January 1990. For emerging markets, our sample starts in January 1996, due to lack of stock return and foreign exchange data in certain countries. Both samples end in December 2015. Stock returns, stock exchange country codes, country incorporation codes and currency codes for international stocks are taken from the Compustat Global Security Daily file. Information on foreign exchange rates comes from Bloomberg. For developed market countries, we obtain 4 MSCI classifies an equity market by its stage of financial development into three groups: developed, emerging, and frontier. During our sample period, three countries were reclassified by MSCI. Greece was upgraded from emerging to developed in May 2001, and then downgraded back to emerging in November 2013. Israel was upgraded from emerging to developed in May 2010. Portugal was upgraded to developed markets in November 1997. 5

aggregate book-to-market from Ken French s website. 5 We obtain the number of sell-side analysts reporting next-year EPS estimates from I/B/E/S. For many of our tests that follow, we sort stocks into group by firm size. Specifically, we use an aggregate market capitalization breakpoint methodology which is also applied by index providers such as MSCI and other asset managers. Specifically, at the beginning of each year we assign a capitalization score (F k) based on the sum of those stocks market capitalization with the same or lower market capitalization divided by the total market capitalization of the eligible universe (either Developed or Emerging Market stocks) multiplied by 100. A stock s score captures the percentage of aggregate market capitalization of stocks that have lower or equal market capitalization values. For example, a stock with a market capitalization of $2 billion would have a score of 75 if 75% of the total capitalization of the stock market consistents of stocks with market capitalizations of less than $2 billion. F k = k j=1 N j=i Cap j 100 n, j where F k F j Cap i Take for example three stocks, A, B, and C, which have capitalizations of $200 MM, $300 MM, and $500 MM, respectively. The capitalization score for stock A is equal to 200/(200 + 300 + 500) 100 = 20, and the score for security B is equal to (200 + 300)/(200 + 300 + 500) 100 = 50. The score of 50 for stock B indicates that 50% of the aggregate market capitalization (including stock B) has the same or a lower market capitalization, while 50% of the market has a higher market capitalization. Using each stock s capitalization score, which is calculated annually at the end of December, we put stocks into three groups: Large (F k 30), Mid (30>F k 15) and Small (15>F k 1). We exclude the bottom 1% of stocks, which consist of Microcap stocks that are likely to be illiquid and hard to trade. While this is similar to the methology that MSCI uses, we depart in one key way instead of dividing stocks into groups by country or region breakpoints, we define the breakpoints across the entire universe of developed or emerging market stocks. Thus, even though Great 5 For more information, see Ken French s website: http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html 6

Britain has larger firms on average, its mid-cap stocks are similar in size to mid-cap stocks from other countries. In this way, our breakpoints are homogenous across the respective developed and emerging universes. Exhibit 1 displays index weights for multi-country portfolios that are weighted by end-ofyear market capitalization. Each chart reports (i) the largest country s weight, (ii) the sum of the next four largest countries weights, and (iii) the sum of the remaining countries weights. The top chart shows weight distributions for developed markets; the bottom chart illustrates results for emerging markets. The following 19 countries were classified as developed: Australia, Austria, Belgium, Denmark, Finland, France, Germany, Great Britain, Hong Kong, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Singapore, Spain, Sweden, Switzerland; while 18 markets were classified as emerging: Brazil, Chile, Greece, Hungary, India, Indonesia, Israel, Malaysia, Mexico, Peru, Philippines, Poland, Portugal, South Africa, South Korea, Taiwan, Thailand and Turkey. [Insert Exhibit 1 Here] Investing in foreign markets provides exposures to different economic (GDP growth, inflation, unemployment, and industrial production) trends, which can provide diversification benefits to a portfolio that currently only invests in domestic assets. Exhibit 1 illustrates one of the major challenges associated with using a capitalization-weighted passive index to gain international exposure: in most cases, only a few countries make up most of the index. For developed countries displayed in the top graph, Japan is the largest country, representing 67.2% of the developed sample at the end of 1989. While Japan is still the largest country at the end of the period, its weight had dropped to 18.4% of developed markets at the end of 2015. The trend is for the larger countries to represent less of the portfolio over time. Despite this decline, the five largest countries (black plus shaded line regions) still represent 63% of the developed markets at the end of the sample period. The bottom graph displays results for emerging markets and also shows a decline, with 68% of the index initially being represented by the following five markets (ranked descending by 7

aggregate market capitalization): Thailand, Malaysia, South Africa, Taiwan, South Korea at the end of 1995, while the top five countries, India, South Korea, Taiwan, Thailand and South Africa still represent 65% of the Index at the end of 2015. Exhibit 1 suggests more generally that while capitalization-weighted passive international indices invest in multiple countries, the actual economic exposure is concentrated in only a few large countries due to those countries being much larger in terms of market size when compared to the other, smaller countries. Appendix Exhibit 1 provides country codes for developed and emerging market countries ranked on aggregate market capitalization used in this study. The country rankings in our paper are different from rankings based simply on total country market capitalization due to exclusion of stocks that are incorporated but trade on a different exchange (such as an ADR or GDR). These restrictions particularly reduce the aggregate market capitalization of those countries like Brazil that have historically had large corporations such as Petrobras and Vale that trade on larger international stock exchanges. Despite these exclusions, our value-weight country index returns generally have very high correlations (>95%) with reported MSCI index returns. Do large countries that make up more of a passive index have lower average returns when compared to smaller countries? Exhibit 2 illustrates the growth of wealth by investing in a valueweighted portfolio consisting of stocks for the same groupings of countries (Largest Country, Next Four Largest Countries, and Rest of the Market) as Exhibit 1. The first graph reports results for Developed Markets and shows that a $1,000 investment in the largest country made on January 1, 1990 grows to only $1,015 by December 31, 2015. In contrast, investing in the next four largest countries (which often includes Great Britain, France and Germany) yields $5,108, and investing in the remaining fourteen countries generates $9,227 at the end of the sample period. Our results are consistent with previous studies. [Insert Exhibit 2 Here] The second graph in Exhibit 2 presents results for Emerging Markets and finds that $1,000 invested in January 1, 1996 declines to $92 for an investment in the largest country, $3,037 for the next four largest countries and $4,961 for the remaining thirteen countries. The extremely poor returns associated with an investment in the largest country for the emerging market sample 8

requires some explanation. First, the starting point for the sample is just before the Asian Financial Crisis (1997), which adversely affected many of the larger countries in Emerging Markets, including Indonesia, Malaysia, Russia, South Korea, Taiwan, and Thailand 6. For example, Thailand which was the largest market in December 1995, declined 36% in 1996. As a result of that decline, Malaysia became the largest Emerging Markets country in December 1996, and subsequently experienced a drop of 68% in 1997. In contrast, an investment in the thirteen smallest markets increased 12% in 1996 and 2% in 1997. II. Do Small Country Indices Have Higher Average Returns Than Large Country Indices? Exhibit 3 reports performance summary statistics for different countries in Developed Markets (Panel A) and Emerging Markets (Panel B). The left-hand side of the exhibit presents summary statistics on the country-level, while the right-hand side shows results on the firm-level. The countries are ranked by aggregate market capitalization as of the start of the sample period. The five largest developed market countries have average annual returns of 5.2% per year, compared to 7.8% for the other fourteen markets. The fourteen smaller markets have average annual volatility of 21.7%, which is slightly higher than the 18.6% for the largest five markets. The higher volatility for smaller countries is likely due to having fewer firms listed on their markets, which reduces diversification. Despite the slightly higher volatility, the smallest markets have high risk-adjusted returns (Average Sharpe Ratio of 0.34), which is 50% higher than the performance for the largest five markets (Average Sharpe Ratio of 0.23). Smaller markets tend to have slightly higher betas of 0.97 (measured against the MSCI World Index), compared to an average of 0.92 for the five largest markets. 7 The five largest developed markets have slightly lower average stock volatility (31.3%) when compared to the smaller fourteen markets (32.1%). We also find slightly higher stock-level betas for stocks from smaller markets (0.80 compared to 0.75 for the five largest markets). Larger 6 For more information on the 1997 Asian financial crisis please see https://en.wikipedia.org/wiki/1997_asian_financial_crisis. 7 For country-level summary statistics, we report compound annual returns, volatilities and betas with respect to the MSCI World Index using value-weighted country index monthly returns. The Sharpe Ratio is the average monthly return less the monthly risk-free rate (US one-month treasury rate) divided by the standard deviation of the difference between the monthly country return less the risk-free rate multiplied by the square root of 12. 9

markets tend to have larger firms (13.6 BN geometric-weighted average market capitalization) when compared to the fourteen smaller markets (9.4 BN) and a much greater average number of firms (732 average number of firms each year for the five largest markets compared to 104 firms for the remaining, smaller markets). 8 There are, however, some exceptions, such as Finland, which is the fourth smallest developed country but has an average market capitalization of 12.5 BN due to Nokia, an information technology firm with a nearly $40 BN market capitalization as of the end of December 2015. [Insert Exhibit 3 Here] Panel B reports results for the emerging market sample and shows a similar picture as Panel A. The five largest markets have average annual returns of 2.4% per year, with volatility of 27.2% and a Sharpe Ratio of 0.146. The smaller thirteen markets have average annual returns of 6.2% per year, with 29.0% in volatility and an average Sharpe Ratio of 0.29. We also find that larger emerging markets have slightly lower index betas, higher firm volatilities, higher stock-level betas, higher average firm size and many more firms when compared to smaller emerging markets. The last two rows of each Panel present summary statistics for combinations of developed and emerging markets country indices. The second-to-last row reports value-weighted results that weight countries by market capitalization as with a passive index. The last row reports equalweight results that weight countries equally (the country-indices themselves, however, are valueweighed). As we show, the return associated with an equal-weight portfolio of country indices (8.1%) is nearly twice as big as that of the value-weighted portfolio (4.3%), which has a much greater weight for larger countries. Interestingly, the volatility of the equal-weighted developed market portfolio (16.2%) is actually slightly lower than for the value-weighted portfolio (16.6%), as the benefits of diversifying across countries outweighs the slighty higher volatility and beta associated with smaller countries. The result of reducing weight of large countries and reallocating weight to smaller countries results in a 140% increase in risk-adjusted returns as measured by the Sharpe Ratio. For Emerging Markets in Panel B, the equal-weight portfolio has average returns of 8 For stock-level summary statistics, we report time-series averages of volatilities and betas with respect to the MSCI World Index based on daily returns over the prior year. 10

7.5% and volatility of 20.2% per year, resulting in a Sharpe Ratio of 0.35. In contrast, the valueweighted portfolio yields returns of 3.9%, volatility of 20.5%, and a Sharpe Ratio of only 0.18. We start our analysis by examining Fama-MacBeth regressions of country index returns on different country-level factors. Exhibit 4 forms country indices by capitalization-weighting stocks from a particular country. By examining country indices rather than individual stocks, the methodology in this section is more in line with past literature that originally identified the country effect. We perform this exercise including all stocks (All Cap) and also excluding the largest stocks (Mid/Small). Exhibit 4 is organized as follows. We first run univariate regressions of four different factors natural log of country size, natural log of average firm size within a country, natural log of aggregate country book-to-market, and past 12-month country momentum. We then regress country index returns on all four factors and repeat this exercise for country indices formed from only mid- and small-capitalization firms. 9 [Insert Exhibit 4 Here] The univariate regressions show that country size Ln (Ctry) and average firm size LN (Avg. Firm Size) are negatively related to average stock returns, while book-to-market Ln (Ctry B/M) and momentum Ctry MOM are positively related to average stock returns. Our variable of interest, country size, has a t-statistic that is close to or greater than 2 in each of the univariate regressions (reported as the first regression in each set of five regressions). While average firm size is only significant for Emerging Markets, country momentum is only significant for developed markets and country aggregate book-to-market ratio is not significant for the Developed Market sample. The last two regressions consider all four factors together in a multi-variate setting. As we show, for All Countries (ALL) and Developed Markets (DM), we find coefficients for country size of -0.14 to -0.16 with t-statistics that range from 1.91 to 2.58, respectively. There is not much difference between using all stocks when forming country indices or using only mid- and smallcapitalization stocks. In contrast, the average firm size and book-to-market ratio s coefficients are insignificant for the ALL and DM samples. For the ALL sample, the coefficient on momentum is 9 We repeated our univariate regressions using only mid- and small-capitalization stocks to form value-weighted country indices and found similar (unreported) results to using all stocks to form country indices. 11

significant, but only for the ALL country sample, which uses only Mid/Small Capitalization stocks when forming country indices. In Emerging Markets, we find that adding average firm size and country momentum causes country size to be an insignificant predictor of average returns. We find some evidence of average firm size predicting negative future returns after controlling for country momentum and country size, but only when large stocks are included this variable also becomes much weaker when we only use Mid/Small stocks when forming country indices. Multicollinearity is a bigger issue in Emerging Markets, as small countries also tend to have had positive momentum and tend to have smaller average firm size. III. Do Stocks in Small Countries Have Higher Returns than Stocks in Large Countries? In this section we change focus slightly and examine the returns of individual stocks rather than those of indices. The advantage of examining individual stocks is that we can explicitly separate the effect of country size and firm size. In addition, we can examine the interaction between the small country effect and firm size. Recall that we conjectured that because home bias affects small stocks more than large stocks,the small country effect will be stronger for smaller cap stocks. Exhibit 5 reports value-weighted monthly returns for stocks sorted on firm size (Large, Mid and Small) and country size (Largest Country, Next four Largest Countries and Remaining Countries). As we show in Panel A, the largest country has very poor performance, ranging from 0.09% - 0.25% for Developed Markets. In contrast, average performance for the Next Four Largest Countries and Remaining Countries is 0.70% and 0.90%, respectively. Panel B presents results for Emerging Markets and finds average returns for the Next Four Largest Countries is 0.76%, while for Remaining Countries the average return is 0.92%. The returns for investing in the largest country in Emerging Markets is abysmal, ranging from -0.76% for small capitalization stocks to - 0.62% for large capitalization stocks, compared to an average return of 0.57% for the Next Four Largest Countries and 0.69% for the Remaining Countries. Our results presented in Exhibit 5 suggest that the country size premium is largely independent of firm size. [Insert Exhibit 5 Here] 12

A major difference between the analysis in this section relative to the analyses presented in Exhibit 4 is that by using index returns rather than individual stock returns, we are effectively weighting each country equally. Our main analysis is presented in Exhibit 6, which reports results from regressions of stock returns on the natural log of firm size, country size and a dummy for emerging markets. Our methodology involves using individual stock returns, which enables us to control for firm size when examining whether country size explains stock returns. The first set of four regressions shows that within all size groupings, the log of Country Size is a negative and significant predictor of future stock returns. We see a slightly stronger country size effect among Large and Mid-Capitalization stocks, and contrary to the early US evidence we do not find evidence of a small firm effect. Emerging markets also have under-performed Developed Markets during our sample period, leading to a negative relation between the EM Dummy and average stock returns. [Insert Exhibit 6 Here] The second set of regressions examines all stock markets but excludes Japan, which is the biggest market in our sample. Dropping Japan reduces the sample by roughly 25% across all size groups. While the Ln (Ctry) coefficient drops in each regression, we still find that country size is a significant negative predictor of future stock returns. For the non-large size groupings, we also find a significant positive coefficient on firm size. The third and fourth sets of regressions divide stocks by Developed and Emerging Markets. As we show, the natural log of country size is negatively related to average stock returns in both tests, with slightly stronger results in Emerging Markets. The last two sets of regressions split the sample by time period. Note that in the earlier time period prior to 2003, there are far fewer small stocks when compared to the sample after 2002. For the earlier sample period, regression coefficients are roughly two to three times as large as in the later sample period. Also, the t-statistics for the coefficient on Ln (Ctry) are not significantly different from zero for the regressions using the later time period between 2003 and 2015. These results address a number of issues that were raised in the introduction. First, we show that the country size effect is largely independent of the firm size effect and is a significant negative predictor of future returns among groups of firms with different sizes. Second, we find 13

that our country size results persist even if we exclude the largest market, Japan. Third, this is not an emerging markets effect; we observe a country size effect in both emerging and developed countries. Recall that our main explanation involved an interaction between home bias and market size. Our results thusfar are mixed on whether this interpretation explains the country size effect. For example, the lack of capital market access for firms in small markets would suggest these stocks are riskier consistent with the hypothesis that there is weak evidence that smaller countries have higher volatilities and betas to the world index. Second, small stocks (relative to large stocks) should be more sensitive to the country size effect. We find the country size effect exists among different sized stock groups, and in certain circumstances is stronger among large stocks, which is inconsistent with this explanation. Last, we should expect that over time, globalization forces and foreign flows into smaller markets should improve capital market access for firms in small markets. As we show, our results supporting this premise are in fact stronger in the earlier part of the sample period. IV. Do Small Countries Attract Less Analyst Coverage? In this section we directly examine whether smaller countries are more subject to home bias by looking at analyst coverage. Our conjecture is that there is a fixed cost associated with covering the stocks in a particular country, and because of those costs investors and analysts may choose to ignore the smaller countries. If this is the case, then when we control for firm size and industry we should observe less coverage of stocks in smaller countries. Exhibit 7 reports annual panel regressions of the natural log of one plus the number of analysts covering a stock on firm and country size. We calculate our measure once a year on January 1 st, using the number of analysts that report EPS estimates in December of the prior year. We control for year and industry fixed effects and the natural log of firm size, LN(Firm). We also cluster errors by firm and time period. [Insert Exhibit 7 Here] The first four panel regressions in Exhibit 7 report results for developed markets across different sized firms (All Cap, Large, Mid and Small Cap); the second set of regressions reports results for emerging markets. The All Cap and Small Cap results are similar, as close to 80% of 14

firms in the All Cap developed market sample are Small Cap firms. With the exception of large firms in developed markets, our regression results show that bigger firms are significantly more likely to have more analysts following those stocks and that much of the explanatory power in these regressions comes from firm size. Our main variable of interest is the natural log of country size, Ln (Ctry). As we show, All and Small-Cap developed market stocks have a significant negative relation between country size and analyst following. This result is inconsistent with our conjecture that analyst coverage should be less in smaller countries; however, it should be noted that this result is driven by small cap stocks that have very little analyst coverage. We do find a significant positive relation between country size and analyst following among developed market large capitalization stocks, which is consistent with our conjecture that smaller country stocks receive less attention for international investors. As we expect, the positive relation between large capitalization coverage and country size is especially strong in the emerging markets. V. Conclusion This paper takes a closer look at the country size premium; i.e., the tendency of stocks from smaller markets to have higher returns than stocks in the largest markets. Our measure of country size is the sum of all stocks market capitalization within a particular country. We find that the country size effect is largerly independent of the firm size effect, exists when excluding the largest country (Japan), and potentially subsumes other country-level quantitative factors such as value and momentum. Our working hypothesis was that the small country effect was due to home bias that depressed small country stock prices more than large country stock prices because of the small countries more limited investor base. We presented evidence that is consistent with this conjecture in particular, the small country effect seems to be declining as access to international markets improves; and that it is stronger in emerging markets than in developed markets. We were somewhat surprised, however, that the small country effect is as strong for large capitalization stocks as it is for small capitalization stocks. Our analysis of analyst coverage suggests a potential explanation for this phenomenon. We find that for large stocks, analyst coverage is in fact somewhat less in smaller countries, suggesting that at least historically, home bias affected even larger stocks in small countries. 15

References Angelidis, Timotheos and Nikolaos Tessaromatis. "Global Style Portfolios Based on Country Indices." MPRA Working Paper (2016). Asness, Clifford S., John M. Liew and Ross L. Stevens. "Parallels between the Cross-Sectional Predictability of Stock and Country Returns." The Journal of Portfolio Management, Vol. 23, No.3 (1997), pp. 79-87. Asness, Clifford S., Roni Israelov, and John M. Liew. "International Diversification Works (Eventually)." Financial Analysts Journal, Vol. 67, No.3 (2011), pp. 24-38. Baele, Lieven, and Koen Inghelbrecht. "Time-Varying Integration and International Diversification Strategies." Journal of Empirical Finance, Vol. 16, No.3 (2009), pp. 368-387. Banz, Rolf W. "The Relationship between Return and Market Value of Common Stocks." Journal of Financial Economics, Vol.9, No.1 (1981), PP. 3-18. Braymen, Charles, and Robert R. Johnson. "International Diversification: The Weighting is the Hardest Part." The Journal of Portfolio Management, Vol.42, No.1 (2015), pp. 53-62. Chan, Kalok, Allaudeen Hameed, and Wilson Tong. "Profitability of momentum stragegies in the international equity market." Journal of financial and quantitative analysis 35.02 (2000): 153-172. Christoffersen, P., V. Errunza, K. Jacobs and H. Langlois, Is the Potential for International Diversification Disappearing? working paper, Rotman School of Management, University of Toronto (2011). Eun, Cheol S., and Jinsoo Lee. "Mean Variance Convergence around the World." Journal of Banking & Finance, Vol. 34, No.4 (2010), pp. 856-870. Goetzmann, William N., Lingfeng Li and K. Geert. Rouwenhorst. "Long-Term Global Market Correlations." Journal of Business, Vol.78, No.1 (2005), pp. 1-38. Keppler, Michael and Heydon D. Traub. "The Small-Country Effect: Small Markets Beat Large Markets." The Journal of Investing, Vol.2, No.3 (1993), pp. 17-24. Keppler, Michael and Peter Encinosa. "The Small-Country Effect Revisited." The Journal of Investing, Vol.20, No.4 (2011), pp. 99-103. La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert W. Vishny. "Legal Determinants of External Finance." Journal of Finance (1997), pp. 1131-1150. Li, Tianchuan and Mahesh Pritamani. "Country Size and Country Momentum Effects in Emerging and Frontier Markets." The Journal of Investing, Vol.24, No.1 (2015), pp. 102-108. Santis, Giorgio and Bruno Gerard. "International Asset Pricing and Portfolio Diversification with Time Varying Risk." The Journal of Finance, Vol.52, No.5 (1997), pp. 1881-1912. 16

Sharpe, William F. "The Sharpe Ratio." The Journal of Portfolio Management, Vol.21, No.1 (1994), pp. 49-58. Statman, Meir and Jonathan Scheidb. "Global Diversification. Journal of Investment Management, Vol.3, No.1 (2005), pp. 1-11. 17

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Exhibit 1. Country Weights for Developed and Emerging Market Multi-Country Indices. For each chart, we report the largest country, next four largest countries and remaining countries aggregate portfolio weight based on a capitalization-weighted index using end-of-year country aggregate market capitalizations. The developed markets sample begins in 1990; the emerging markets sample begins in 1996. Both samples end in 2015. Developed Markets 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Largest Country Next Four Largest Countries Rest of the Market Emerging Markets 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Largest Country Next Four Largest Countries Rest of the Market 18

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 1990 1990 1991 1992 1993 1993 1994 1995 1996 1996 1997 1998 1999 1999 2000 2001 2002 2002 2003 2004 2005 2005 2006 2007 2008 2008 2009 2010 2011 2011 2012 2013 2014 2014 2015 Exhibit 2. Growth in Wealth of $1,000 Invested in Different-sized Financial Markets. For each chart, we report the growth in wealth associated with investing in a value-weighted portfolio consisting of the largest country, next four largest countries, and remaining countries using end-of-year country aggregate market capitalizations, rebalanced on January 1 of each year. The developed markets sample begins in January 1990; the emerging markets sample begins in January 1996. Both samples end in December 2015. Developed Markets 12000 10000 8000 6000 4000 2000 0 Rest of the Market Next Four Largest Countries Largest Country Emerging Markets 7000 6000 5000 4000 3000 2000 1000 0 Rest of the Market Next Four Largest Countries Largest Country 19

Exhibit 3. Summary Statistics by Country. For each country, we report summary statistics based on country indices (left-hand side) and individual stocks (right-hand side). The country characteristics include compound annual return (Return), annualized volatility (Volatility), Beta (Beta to MSCI World) and annualized Sharpe Ratio. The firm characteristics include annualized volatility (Avg. Firm Volatility), beta (Avg. Firm Beta), median average firm market capitalization in billions (Median Cap ($BN)), geometric average firm market capitalization in billions (Avg. Cap ($BN)) and the average number of firms (N). Stocklevel annualized volatility and beta are based on value-weighted averages of daily returns over the previous year. Countries are sorted in descending order, with the largest country by aggregate market capitalization as of the beginning of the sample period listed at the top. Panel A displays results for Developed Market countries from January 1990 to December 2015. Panel B displays results for Emerging Market countries from January 1996 to December 2015. The two rows of each panel display summary statistics for valueweighted (VW) and equal-weighted (EW) portfolios that consist of country indices from all countries in either Developed or Emerging Markets. Panel A. Developed Markets (January 1990 - December 2015) Country Return Volatility Country-Level Characteristics Beta to MSCI World Sharpe Ratio Avg. Firm Volatility Firm-Level Characteristics Avg. Firm Beta Median Cap ($BN) Avg. Cap ($BN) Japan 0.1 20.2 0.87-0.04 36.0 0.62 0.43 7.5 1988 Great Britain 7.7 15.1 0.85 0.37 29.8 0.72 0.51 21.5 676 Germany 3.9 21.4 1.02 0.16 31.6 0.69 0.57 15.0 307 France 5.8 18.7 1.00 0.24 31.0 0.76 0.57 16.9 331 Australia 8.8 17.7 0.84 0.40 28.3 0.96 0.39 7.0 315 Switzerland 10.3 16.8 0.86 0.50 25.8 0.74 0.56 22.1 159 Netherlands 9.1 18.7 0.97 0.41 29.0 0.89 1.09 18.5 112 Spain 6.8 22.0 1.07 0.28 29.7 0.84 1.09 13.7 110 Italy 2.0 23.0 1.03 0.08 32.3 0.90 0.58 9.5 169 Sweden 9.7 23.5 1.20 0.39 34.3 0.85 0.56 6.2 149 Belgium 11.3 18.5 0.82 0.52 28.9 0.66 0.52 10.6 86 Hong Kong 11.7 25.3 1.01 0.45 33.4 0.98 0.68 14.8 114 Singapore 7.0 23.0 1.02 0.28 32.0 0.95 0.35 3.5 169 Denmark 11.3 18.1 0.84 0.53 30.7 0.64 0.47 4.9 78 Norway 7.7 22.4 0.99 0.32 34.6 0.83 0.37 5.1 84 Finland 4.0 28.8 1.19 0.18 41.0 0.80 0.54 12.5 76 Austria 2.9 22.8 0.91 0.12 33.1 0.65 0.55 2.4 51 Ireland 7.0 22.2 1.04 0.29 37.7 0.82 0.74 6.5 51 New Zealand 8.8 18.3 0.69 0.39 27.1 0.65 0.32 1.2 44 Dev. Mkts(VW) 4.3 16.6 1.00 0.16 31.9 0.76 0.48 11.3 5067 Dev. Mkts(EW) 8.1 16.2 0.96 0.38 31.9 0.76 0.48 11.3 5067 N 20

Panel B. Emerging Markets (January 1996 - December 2015) Country-Level Characteristics Firm-Level Characteristics Country Return Volatility Beta to MSCI World Sharpe Ratio Avg. Firm Volatility Avg. Firm Beta Median Cap ($BN) Avg. Cap ($BN) Thailand 0.1 30.2 1.03 0.08 44.5 0.68 0.14 1.6 182 Malaysia 2.0 26.0 0.76 0.12 33.1 0.79 0.12 2.0 252 South Africa 6.3 23.9 1.01 0.28 35.3 0.78 0.33 4.4 153 Taiwan 1.7 25.9 0.94 0.10 36.5 0.84 0.17 3.0 460 Korea 2.4 30.1 1.20 0.15 46.9 0.92 0.09 4.4 367 India 8.5 29.0 0.97 0.35 42.8 0.81 0.15 4.0 339 Indonesia 5.8 38.0 1.29 0.28 50.2 0.92 0.17 2.6 108 Mexico 12.9 23.4 1.09 0.54 33.6 0.82 0.68 7.4 76 Phillipines 2.3 26.7 0.90 0.13 38.9 0.81 0.18 1.7 66 Brazil 5.1 32.3 1.19 0.25 58.0 1.01 0.57 2.2 26 Chile 6.3 21.2 0.84 0.29 26.6 0.64 0.36 2.9 81 Israel 8.9 22.1 0.87 0.39 33.4 0.72 0.15 2.3 111 Turkey 5.7 39.6 1.64 0.32 52.8 1.27 0.16 2.3 108 Portugal 3.1 21.9 0.89 0.15 31.8 0.70 0.41 4.1 31 Greece -3.0 34.2 1.27 0.01 47.2 1.01 0.22 2.0 73 Peru 9.2 22.7 0.65 0.40 34.8 0.43 0.20 1.3 30 Poland 4.6 32.0 1.35 0.23 38.1 1.19 0.11 2.0 80 Hungary 10.8 33.4 1.44 0.41 40.1 0.90 0.21 3.3 21 Emer.Mkts(VW) 3.9 20.5 1.05 0.18 40.3 0.85 0.15 3.3 2564 Dev. Mkt(EW) 7.5 20.2 1.07 0.35 40.3 0.85 0.15 3.3 2564 N 21

Exhibit 4. Fama-MacBeth Regressions of Monthly Country Index Returns on Country Size, Aggregate Country Book-to-Market and Past One-Year Momentum for Developed and Emerging Market Stocks. This table reports the results of a set of Fama-MacBeth regressions of monthly country returns on country size, country average firm size, aggregate country book-to-market and countrylevel momentum. EM/DM refers to whether the regressions only use emerging market stocks (EM) or developed market stocks (DM). Time period reflects the starting and ending year for each regression. N is the average number of firms or countries in the sample each year. Ln (Ctry) is the natural log of the aggregate market capitalization of a country measured as of the end of December of the previous year. Ln (Firm) is the natural log of the geometric average firm size for stocks from a specific country measured as of the end of December of the previous year. LN (B/MCtry) is the natural log of the ratio of aggregate book equity divided by aggregate market capitalization for stocks from a specific country measured as of the end of December of the previous year. Ctry MOM is the past one-year country value-weighted return. For the sake of brevity, the intercept is not reported. T-statistics are reported in parentheses to the right of each estimate and are based on Newey-West corrected standard errors with a lag of 12 months. EM/DM Size Ln (Ctry) t-stat Ln (Firm) t-stat Ln (B/M Ctry) t-stat Ctry MOM t-stat EM Dummy t-stat N Time Period ALL All Cap -0.17 (2.66) -0.22 (0.99) 32 1991-2015 ALL All Cap -0.12 (1.56) -0.18 (0.86) 32 1991-2015 ALL All Cap 0.92 (1.81) 0.08 (0.45) 32 1991-2015 ALL All Cap -0.14 (1.91) 0.06 (0.53) 0.78 (1.64) -0.15 (0.83) 32 1991-2015 ALL Mid/Small -0.14 (1.92) 0.05 (0.45) 1.29 (2.52) -0.09 (0.51) 32 1991-2015 DM All Cap -0.12 (2.23) 19 1991-2015 DM All Cap -0.03 (0.41) 19 1991-2015 DM All Cap 0.45 (1.70) 19 1991-2015 DM All Cap 1.25 (2.33) 19 1991-2015 DM All Cap -0.16 (2.58) 0.13 (1.22) 0.32 (1.46) 0.84 (1.33) 19 1991-2015 DM Mid/Small -0.14 (2.05) 0.08 (0.87) 0.21 (0.75) 1.03 (1.60) 19 1991-2015 EM All Cap -0.28 (1.97) 18 1997-2015 EM All Cap -0.36 (2.40) 18 1997-2015 EM All Cap 0.47 (0.69) 18 1997-2015 EM All Cap -0.05 (0.40) -0.33 (1.99) -0.05 (0.70) 18 1997-2015 EM Mid/Small -0.10 (0.80) -0.22 (1.06) 1.15 (1.60) 18 1997-2015 22

Exhibit 5. Value-weighted Monthly Returns for Different-sized Financial Markets by Firm Size Group. At the beginning of each year, stocks are sorted into three groups according to market capitalization: Large (Top 70% of aggregate capitalization), Mid (70% to 85%), and Small (85% to 99%). Within these three groups based on size, stocks are further divided into three groups: (i) largest country, (ii) next four largest countries, and (iii) remaining countries using end-of-year country aggregate market capitalizations, rebalanced on January 1 of each year. We form value-weighted portfolios comprised of stocks in these various groups. The rankings of countries are dynamic and change each year. Panel A displays developed market results starting in January 1990; Panel B reports emerging market results starting in January 1996. Both sample periods end in December 2015. Panel A. Developed Markets (January 1990 December 2015) Large Mid Small 0.62% 0.81% 0.72% 1.01% 0.77% 0.88% 0.09% 0.22% 0.25% Largest Country Next 4 Largest Countries Remaining Countries Largest Country Next 4 Largest Countries Remaining Countries Largest Country Next 4 Largest Countries Remaining Countries Panel B. Emerging Markets (January 1996 December 2015) Large 0.61% 0.80% Mid 0.93% 1.01% Small 0.75% 0.95% -0.62% Largest Country Next 4 Largest Countries Remaining Countries -0.43% Largest Country Next 4 Largest Countries Remaining Countries -0.76% Largest Country Next 4 Largest Countries Remaining Countries 23