On-line Appendix: The Mutual Fund Holdings Database

Transcription

1 Unexploited Gains from International Diversification: Patterns of Portfolio Holdings around the World Tatiana Didier, Roberto Rigobon, and Sergio L. Schmukler Review of Economics and Statistics, forthcoming On-line Appendix: The Mutual Fund Holdings Database This on-line appendix provides more details on how we construct the mutual fund portfolio holdings database. Our data source is Morningstar International Equity Mutual, a private company that collects mutual fund data. Our sample consists of the universe of actively managed open-ended U.S. mutual funds established to purchase equity assets around the world. We exclude from the analysis the funds that invest solely in bonds and the funds that invest in both bonds and equity. We also exclude the funds that focus on a specific sector (such as metals, health care, and utilities). Some of the mutual fund families in our database sell the same portfolio to investors under slightly different names, depending on their fee structure and minimum investment requirements. In this paper, we consider these different funds with identical portfolios only once (we do not treat them as separate funds). For example, Fidelity Advisors has the following Latin America funds, all of which hold the same portfolio: Fidelity Advisors Latin America A, Fidelity Advisors Latin America B, and Fidelity Advisors Latin America T. We only use one portfolio to represent these funds and call it Fidelity Advisors Latin America. We analyze the monthly portfolio reports from March 1992 until June Despite the availability of these monthly reports, most mutual funds only disclose the updated portfolio holdings on an annual, semi-annual, or quarterly basis, depending on the U.S. Securities and Exchange Commission (SEC) reporting rules at the time and on their own agreement with Morningstar. These updated portfolio reports are not published on the same months by all funds. Given this heterogeneity in the release of new information by mutual funds, we construct our database with the last reported portfolio information for each fund in any given year. For example, our sample of mutual fund holdings for 2005 contains the portfolio data for the Fidelity Worldwide Fund as of October 2005 and the portfolio data for the Scudder Fund as of December We use the following information for each asset included in the mutual fund portfolios, as reported by the funds themselves: the stock name, the industry classification, the amount invested in each stock by each fund, the country of origin, and the firm identifiers (if available). A difficulty in constructing this portfolio holding database is that mutual funds disclose their asset allocation in separate reports over time. Their holdings are not linked across reports and firm names are not consistently reported over time either. There is also heterogeneity in the information reported across funds. Homogenizing and matching the information on this dataset is thus not a simple task because stock identifiers are rarely available and, when present, are not always unique. For example, stock identifiers typically change with the location where they are traded. 1

2 For the analysis in the paper, we need information on the country of origin of the assets in mutual fund portfolios to assess the cross-country allocations across funds over time. We also need to match the asset-level holdings across funds within mutual fund families in order to evaluate the similarity in fund portfolios. Lastly, we need to match the portfolio holdings over time to construct the portfolio turnover statistics. Overall, our approach consists in assembling a master list of firms with unique identifiers based on mutual fund reports that allow us to consistently match our mutual fund holdings database across funds over time. In constructing this master list of firm identifiers, we take into account the different spellings of firm names and we assign a single identifier. For example, we assign the same identifier to the mutual fund holdings on Companhia Energetica de Minas Gerais (Brazil), which are reported under a wide range of names (e.g. CEMIG, CEMIG (144A), CEMIG ADR, CEMIG Pfd, Cemig Cia Energ Mg, Cia Energetica de Minas Gerais, among many others). We also take into account changes in company names over time. For example, Union Petrochemical Corp. (Taiwan) changes its name to UPC Technology Corp. (Taiwan) in June We assign the same firm identifier to portfolio holdings on assets of both companies. A key piece of information for our matching is the country of origin of the portfolio holdings, which is available in the mutual fund reports. However, this information is only available for the period and we do not attempt any matching of holdings for the pre-1997 period. Hence, the analysis in the paper that requires country-level or asset-level matching starts in We also do not attempt to match portfolio holdings when a fund does not report the country of origin nor the identifier of its stock holdings. Lastly, we do not match the U.S. holdings. We make one further adjustment to the asset-level database based on the mutual fund loadings on each stock. We exclude from the analysis all of the fund-year observations for which the sum of the loadings on each stock (by a given fund in a given year) exceeds by 10% or more the mutual fund self-reported total assets under management. These observations are excluded due to conflicting information reported by the mutual funds themselves. We exclude 8,224 observations, or 0.89% of the matched sample. Note that not all the analysis in the paper based on the holding database requires this asset-level matching information. For instance, the analysis on the number of holdings and the loadings in the top ten assets (available since 1991) does not require any sort of matching of portfolio holdings. In the notes to each table and figure presented in the paper, we explicitly mention the sample used to obtain the statistics shown. The on-line Appendix Table 1 describes our final holding dataset. This dataset comprises the end-of-year detailed information on portfolio holdings between 1991 and 2005, covering 499 mutual fund families and a total of 1,904 funds, with a total of 8,420 fund-year observations. Each mutual fund family has on average four different mutual funds. Our dataset has 1,359,750 asset-level holdings for all funds in all years in our sample, 926,695 of which are matched. For each matched holding in our dataset, we also use information on the industry classification as reported by mutual funds themselves. There is some variation in the wording used by mutual funds to classify their holdings to the different industries. We homogenize these different 2

3 classifications into the following groups, as shown in the top panel of Appendix Table 2: (i) consumer goods, (ii) financials, (iii) health, (iv) industrial, (v) services, (vi) technology, and (vii) utility. This self-reported sectoral classification allows us to identify the sector of 62% of the matched holdings in our sample, that is, information for 575,044 observations out of the 926,695 matched stock holding observations between 1997 and If a particular fund-holding does not have a sectoral classification identified by the fund itself, we assign the classification most cited by the other funds in our sample, thus allowing us to classify a total of 96% of the matched stock holdings in our sample. 1 We exclude the remaining unclassified stocks from the sectoral analysis reported in Table 3 of the paper. For robustness purposes, we complement this information with that of the Standard Industrial Classification (SIC). The information on the SIC sectoral classification comes from Worldscope and we merge it with our holdings dataset by using the firm names and the country of origin. This SIC sectoral classification allow us to classify 78.5% of the portfolio holdings sample, that is, information for 727,108 observations out of the 926,695 observations on stock holdings between 1997 and SIC codes allow us to distinguish ten different sectors, as shown in the bottom panel of Appendix Table 2: (i) agriculture, forestry, and fishing, (ii) mining, (iii) construction, (iv) manufacturing, (v) transportation, communications, and utilities, (vi) wholesale trade, (vii) retail trade, (viii) finance, insurance, and real estate, (ix) services, and (x) public administration. 1 The mutual fund portfolio holdings need to be matched in order to allow us to use the classification most cited by the other funds in our sample. This is the reason why we work with the sample of matched holdings instead of the entire sample of mutual fund holdings between 1997 and

4 Appendix Figure 1. Cumulative Distribution of the Number of Mutual Fund Holdings across Sectors Panel A. Most Cited Classification of Sectors 1 Median across Specialized 0 S1 S2 S3 S4 S5 S6 S7 Sectors Panel B. SIC Classification of Sectors Median across Specialized S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 Sectors This figure shows separately the median cumulative distribution of the stock holdings across sectors for global and specialized funds over the period. The portfolio allocations are based on the number of stock holdings. In Panel A, the sectoral classification of the mutual fund holdings follows the most cited sectoral classification of each stock by all of the mutual funds in our sample. The stock holdings are thus consistently classified across funds over time. In Panel B, the sectoral classification of the mutual fund holdings follows the SIC classification of sectors.

5 Appendix Figure 2. Cumulative Distribution of the Mutual Fund Assets across Sectors Panel A. Mutual ' Own Sector Classification Median across Panel B. Most Cited Classification of Sectors Specialized S1 S2 S3 S4 S5 S6 S7 Sectors Median across S1 S2 S3 S4 S5 S6 S7 Sectors Panel C. SIC Classification of Sectors Specialized Median across Specialized S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 Sectors This figure shows separately the median cumulative distribution of the mutual fund net assets under management across sectors for global and specialized funds over the period. In Panel A, the sectoral classification of the mutual fund holdings follows the funds' own classification. If a particular fund holding is not classified by the reporting fund, we assign the most cited classification for that asset by other funds in our sample. In Panel B, the sectoral classification of the mutual fund holdings follows the most cited sectoral classification of each stock by all of the mutual funds in our sample. The stock holdings are thus consistently classified across funds over time. In Panel C, the sectoral classification of the mutual fund holdings follows the SIC classification of sectors.

6 Appendix Figure 3. Evolution of the Within-Fund Entropy Measure for Stock Holdings All Stock Holdings Excluding the U.S. Holdings Holdings in Developing Countries Only World Foreign Specialized This figure shows the within-fund entropy measure for stock holdings by fund type from 1997 to The within-fund entropy measure captures the commonality of stock holdings in two consecutive years for each individual mutual fund. The figure reports the median entropy measure across all funds of a certain type in a given year. In the top panel, we use all stock holdings, except the U.S. stocks. In the bottom panel, we use only the holdings in developing countries. Specialized funds comprise emerging market, regional, and country funds.

7 Appendix Figure 4. Dispersion in the Number of Holdings Frequency Histogram of the Number of Mutual Fund Holdings Summary Statistics: Average: 149 Stocks Median: 95 Stocks Std. Dev.: 186 Stocks 73% of Obs. below 150 Holdings 86% of Obs. below 250 Holdings 91% of Obs. below 350 Holdings No. of Holdings This figure shows the dispersion in the number of stock holdings across mutual fund over the period. The histogram for the distribution of fund-level observations is reported.

8 Appendix Table 1. Data Coverage Stock Holdings Database Sample Frequency Annual Number of Families 499 Total Number of 1,904 Total Number of Fund-Year Observations 8,420 Total Number of Asset-Level Holdings 1,359,750 Total Number of Matched Asset-Level Holding 926,695 Return Database Sample September June 2006 Frequency Daily Number of Families 36 Total Number of 371 Total Number of Observations 722,885 This table describes the two datasets analyzed in this paper. The data source for the mutual fund stock holding database is Morningstar International Equity Mutual. The data source for the mutual fund return database is Bloomberg.

9 Appendix Table 2. Classification of Sectors Mutual Fund Reports Sector 1 Sector 2 Sector 3 Sector 4 Sector 5 Sector 6 Sector 7 Consumer Goods, including Durables and Staples Financial Services Health Industrial, including Industrial Material and Industrial Cyclical Services, including Business and Consumer Services, Media, and Retail Technology, including Hardware and Software Utility, including Energy and Telecommunications SIC Classification Sector 1 Sector 2 Sector 3 Sector 4 Sector 5 Sector 6 Sector 7 Sector 8 Sector 9 Sector 10 Agriculture, Forestry, and Fishing Mining Construction Manufacturing Transportation, Communications, and Utilities Wholesale Trade Retail Trade Finance, Insurance, and Real State Services Public Administration This tables shows the composition of the two sectoral classification analyzed. In the top panel, it shows the sectoral classification of the mutual fund holdings following the mutual fund reports. In the bottom panel, it shows the sectoral classification of the mutual fund holdings following the SIC classification.

10 Appendix Table 3. Median Portfolio Allocation of the Number of Mutual Fund Holdings across Sectors Panel A. Most Cited Classification of Sectors No. of Classified Consumer Goods Financial Services Health Industrial Services Technology Utility Holdings (S1) (S2) (S3) (S4) (S5) (S6) (S7) Fund Type 88 14% 20% 6% 16% 16% 8% 17% World (Excl. U.S. Holdings) 63 14% 18% 6% 15% 16% 8% 17% Foreign 96 14% 20% 6% 17% 16% 8% 17% Specialized 69 13% 20% 3% 16% 15% 9% 20% Emerging Market % 20% 2% 19% 13% 10% 23% Asia 56 13% 22% 2% 14% 16% 14% 17% Europe 64 13% 22% 8% 13% 16% 6% 18% Latin America 42 14% 14% 0% 21% 21% 0% 25% Country 75 6% 17% 2% 23% 21% 12% 19% All 81 14% 20% 6% 16% 16% 8% 17% Panel B. SIC Classification of Sectors Agriculture, Forestry, and Fishing Mining Construction Manufacturing Transportation, Communications, and Utilities Wholesale Trade Retail Trade Finance, Insurance, and Real State No. of Classified Services Holdings (S1) (S2) (S3) (S4) (S5) (S6) (S7) (S8) (S9) (S10) Fund Type 67 0% 2% 1% 44% 14% 3% 5% 20% 6% 0% World (Excl. U.S. Holdings) 47 0% 3% 0% 44% 15% 3% 4% 18% 6% 0% Foreign 74 0% 2% 1% 44% 14% 4% 5% 21% 6% 0% Specialized 59 0% 3% 1% 38% 20% 4% 5% 20% 5% 0% Emerging Market 94 0% 5% 2% 37% 22% 4% 4% 19% 4% 0% Asia 51 0% 2% 2% 38% 18% 5% 5% 20% 7% 0% Europe 44 0% 0% 0% 41% 15% 4% 4% 23% 7% 0% Latin America 37 0% 4% 0% 34% 29% 2% 7% 16% 0% 0% Country 84 0% 23% 0% 24% 17% 4% 7% 15% 10% 0% All 63 0% 2% 1% 43% 16% 4% 5% 20% 6% 0% Public Administration The table shows the median portfolio allocation across the different sectors bymutual fund type over the period. The portfolio allocations are based on the number ofstock holdings. In PanelA, the sectoral classification of the mutual fund holdings follows the most cited sectoralclassification ofeach stock by all of the mutual funds in our sample. The stock holdings are thus consistently classified across funds over time. In Panel B, the sectoral classification of the mutual fund holdings follows the SIC classification of sectors.

11 Appendix Table 4. Median Portfolio Allocation of the Mutual Fund Assets across Sectors Panel A. Mutual ' Own Sector Classification Consumer Goods Financial Services Health Industrial Services Technology Utility (S1) (S2) (S3) (S4) (S5) (S6) (S7) Fund Type 14% 21% 8% 14% 18% 7% 14% World (Excl. U.S. Holdings) 15% 19% 8% 14% 19% 9% 13% Foreign 14% 21% 8% 14% 18% 7% 14% Specialized 11% 21% 4% 14% 18% 10% 16% Emerging Market 11% 19% 2% 17% 17% 12% 19% Asia 13% 25% 3% 11% 18% 14% 11% Europe 12% 23% 10% 11% 18% 7% 15% Latin America 12% 12% 1% 19% 33% 1% 20% Country 3% 33% 1% 16% 16% 8% 22% All 14% 21% 7% 14% 18% 8% 14% Panel B. Most Cited Classification of Sectors Consumer Goods Financial Services Health Industrial Services Technology Utility (S1) (S2) (S3) (S4) (S5) (S6) (S7) Fund Type 14% 21% 8% 14% 14% 7% 19% World (Excl. U.S. Holdings) 15% 19% 8% 14% 14% 9% 18% Foreign 14% 21% 8% 14% 13% 7% 19% Specialized 11% 20% 4% 14% 13% 11% 24% Emerging Market 10% 19% 2% 17% 11% 13% 27% Asia 11% 23% 2% 11% 13% 16% 19% Europe 12% 23% 11% 11% 13% 6% 21% Latin America 12% 12% 1% 19% 20% 1% 33% Country 3% 33% 1% 16% 16% 8% 22% All 14% 20% 7% 14% 13% 8% 20% Agriculture, Forestry, and Fishing Mining Construction Manufacturing Panel C. SIC Classification of Sectors Transportation, Communications, and Utilities Wholesale Trade Retail Trade Finance, Insurance, and Real State Public Administration Services (S1) (S2) (S3) (S4) (S5) (S6) (S7) (S8) (S9) (S10) Fund Type 0% 2% 1% 46% 14% 3% 4% 21% 5% 0% World (Excl. U.S. Holdings) 1% 3% 2% 47% 15% 3% 4% 19% 6% 1% Foreign 0% 2% 1% 46% 14% 3% 4% 22% 5% 0% Specialized 1% 4% 1% 39% 21% 3% 4% 20% 5% 1% Emerging Market 0% 5% 1% 39% 25% 3% 3% 18% 3% 1% Asia 1% 3% 2% 37% 17% 5% 7% 21% 6% 0% Europe 1% 2% 2% 43% 15% 3% 4% 25% 5% 0% Latin America 1% 5% 1% 34% 34% 2% 7% 13% 1% 2% Country 0% 21% 0% 14% 19% 2% 6% 32% 5% 0% All 0% 3% 2% 44% 16% 3% 4% 21% 5% 1% The table shows the median portfolio allocation across the different sectors by mutual fund type over the period. The portfolio allocations are based on the mutual fund net assets under management. In Panel A, the sectoral classification of the mutual fund holdings follows the funds' own classification. If a particular fund holding is not classified by the reporting fund, we assign the most cited classification for that asset by other funds in our sample. In Panel B, the sectoral classification of the mutual fund holdings follows the most cited sectoral classification of each stock by all of the mutual funds in our sample. The stock holdings are thus consistently classified across funds over time. In Panel C, the sectoral classification of the mutual fund holdings follows the SIC classification of sectors.

12 Appendix Table 5. Size of Mutual Fund Holdings Value of Total Stock Holdings over Stock Market Capitalization Average Mutual Fund Size (US$ Million) Average Median Std. Dev. Fund Type 0.12% 0.01% 0.75% 899 World 0.18% 0.01% 0.86% 1,320 Foreign 0.11% 0.01% 0.72% 758 Specialized 0.12% 0.02% 0.59% 277 Emerging Market 0.15% 0.02% 0.70% 369 Asia 0.12% 0.01% 0.53% 132 Europe 0.08% 0.01% 0.35% 346 Latin America 0.10% 0.02% 0.47% 144 This table shows the average, the median, and the standard deviation of the value of mutual fund stock holdings as a percentage of each stock's market capitalization over the period across fund types. The average value size of mutual funds, measured in millions of U.S. dollars, is also reported. The data on the stock market capitalization at the firm-level are from Worldscope.

13 Appendix Table 6. Probabilities of Being Held by a Mutual Fund Holdings in Developing Countries Only Specialized Probability of: Probability of: Total Not Being Held Being Held Not Being Held 0% 10% 10% Being Held 75% 13% 89% No Specialized Fund 0% 2% 2% Total 100% 75% 25% [No. of Observations] [92,175] This table shows a two-way frequency count for the mutual fund stock holdings from 1997 to The reported numbers correspond to the joint probability of a stock being held by a specialized fund and a global fund in a given year, conditional on a family having both fund types. Each observation is a family-year-stock observation. The total number of observations is reported in the bracket in the bottom-right cell. When a global fund in a given family-year holds a stock in a country not covered by the specialized funds within that family in that year, this observation is counted in the "No Specialized Fund" row. Stock holdings in developing countries only are considered in the analysis. funds comprise both world and foreign funds. Specialized funds comprise emerging market, regional, and country funds.

14 Appendix Table 7. Portfolio Choice of Mutual : Importance of Family Effects Dependent Variable: Dependent Variable: Number of Stock Holdings Percentage of Net Assets in the Top Ten Holdings Independent Variables Dummy: 1 Manager *** *** [10.800] [0.545] Dummy: 2 Managers *** *** [8.694] [0.664] Dummy: 3 Managers *** *** [19.300] [1.012] Dummy: 4 Managers *** *** [20.100] [2.284] Dummy: 5 Managers *** *** [16.560] [1.100] Dummy: 6 Managers *** *** [31.310] [2.323] Dummy: 7 or More Managers *** *** [28.130] [0.971] Year Dummies Fund Type Dummies Family Dummies No No No No No No No. of Observations R-squared Adjusted R-squared 6,321 6, This table reports the regressions of the number of stock holdings (left column) and the percentage of net assets in the top ten stock holdings (right column) on dummies capturing the number of managers. The sample period for the regressions is Each observation is a fund-year observation. The standard errors are clustered at the family level. The standard deviations are shown in the brackets. The ***, **, and * indicate significance at the one, five, and ten percent levels respectively.

15 Type of Number of Comparisons World 8.71% 12.10% 3.43% 0.88% 0.82% % 10.28% 2.56% 0.91% 0.85% 63 Foreign 5.99% 9.68% 3.79% 0.97% 0.91% % 7.53% 2.58% 0.97% 0.91% 77 Pools of World or Foreign 10.60% 14.88% 4.20% 0.86% 0.85% % 11.66% 4.24% 0.93% 0.89% 24 Total 7.65% 11.32% 3.72% 0.92% 0.87% % 9.18% 2.82% 0.94% 0.89% 164 Type of Average Returns (Per Year) Average Returns (Per Year) Average Difference in Appendix Table 8. Benchmarking: Fund Simulations Simulations Using the Largest Number of Average Difference in Standard Deviation of Simulations Using the Largest Number of Standard Deviation of Minimizing the Variance of Returns Maximizing Expected Returns Number of Comparisons Average Returns (Per Year) Average Returns (Per Year) Simulations Using the Longest Available Sample Average Difference in Simulations Using the Longest Available Sample Average Difference in Standard Deviation of Standard Deviation of World 8.71% 11.96% 3.36% 0.88% 0.78% % 10.13% 2.48% 0.91% 0.82% 63 Foreign 5.99% 10.41% 4.51% 0.97% 0.90% % 8.19% 3.28% 0.97% 0.90% 77 Pools of World or Foreign 10.60% 14.69% 4.00% 0.86% 0.83% % 12.44% 4.94% 0.93% 0.87% 24 Total 7.65% 11.61% 4.02% 0.92% 0.85% % 9.55% 3.22% 0.94% 0.86% 164 Number of Comparisons Number of Comparisons This table shows the differences in the average and the standard deviation of mutual fund returns between the actual and simulated global funds. The results in this table are shown for daily returns. The simulated global funds are constructed from the actual global and specialized funds within the same mutual fund family by using two different procedures. The top panel shows the results from minimizing the variance of returns relative to a benchmark index subject to a restriction on expected returns. The bottom panel shows the results from maximizing expected returns subject to a restriction on the variance of returns relative to a benchmark index. An appropriate benchmark index is used for each simulation. The simulations that use the portfolios with the largest number of specialized funds (the longest time series) for each global fund in each family are reported in the left (right) six columns. The pools of world or foreign funds are simulations that include several world (foreign) funds within the same family but with different investment natures (e.g. value, growth, or blend funds). The portfolio weights are updated every period. The realized returns of the simulated portfolios are calculated out-of-sample. The annualized differences in returns are calculated over the entire sample for each simulation. The averages across simulations are then computed and reported. For more details on this benchmarking exercise, see the working paper version of this paper.