Loughran (1997) contends that the value premium is limited to small stocks. For initial perspective on this issue, we examine variants of VMG (also known as HML), the monthly value-growth return of the three-factor model of Fama and French (1993). We construct VMG by forming six portfolios on size (market capitalization or market cap, which is price times shares outstanding) and book-to-market equity (B/M). Specifically, at the end of each June from 1926 to 2004, we sort NYSE, AMEX (after 1962), and (after 1972) Nasdaq firms with positive book equity into two size groups and three B/M groups. Firms below the NYSE median size are small (S) and those above are big (B). We assign firms to growth (G), neutral (N), and value (V) groups if their B/M is in the bottom 30%, middle 40%, or top 30% of NYSE B/M. The six portfolios, small and big growth (SG and BG), small and big neutral (SN and BN), and small and big value (SV and BV), are the intersections of these sorts. The data sources and calculation of book equity follow Davis, Fama, and French (2000), except that the NYSE sample, which extends back to 1926, now includes firms in any of the Moody's manuals, not just industrials.
The six value-weight size-B/M portfolios are the components of the monthly size and value-growth returns of the Fama–French three-factor model. The size factor, SMB (small minus big), is the average of the monthly returns on the three small stock portfolios minus the average of the returns on the three big stock portfolios,
A. The Value Premium in Small and Big Stock Returns
Table I shows summary statistics for the monthly market excess return, RM-RF (the return on the value-weight portfolio of the NYSE, AMEX, and Nasdaq stocks in our sample minus the 1-month Treasury bill rate), and the SMB, VMG, VMGS, and VMGB returns. Summary statistics for returns on the six size-B/M portfolios that we use to construct SMB and VMG are also shown. The sample period is July 1926 to December 2004 (henceforth, 1926 to 2004), but we also present results for July 1926 to June 1963 and July 1963 to December 2004 (henceforth, 1926 to 1963 and 1963 to 2004). Because July 1963 is the start date of the tests in Fama and French (1992, 1993), the period 1926 to 1963 is out of sample relative to early studies of the value premium.
Table I. Summary Statistics for Monthly Returns on Size and Value Factors and the Size-B/M Portfolios Used to Construct Them At the end of each June from 1926 to 2004, we form six value-weight portfolios, SG, SN, SV, BG, BN, and BV. The portfolios are the intersections of independent sorts of NYSE, AMEX (after 1962), and Nasdaq (after 1972) stocks into two size groups, S (small, firms with the June market cap below the NYSE median) and B (big, market cap above the NYSE median), and three book-to-market equity (B/M) groups, G (growth, firms in the bottom 30% of NYSE B/M), N (neutral, middle 40% of NYSE B/M), and V (value, high 30% of NYSE B/M). Book equity is Compustat's total assets (data item 6), minus liabilities (181), plus balance sheet deferred taxes and investment tax credit (35) if available, minus (as available) liquidation (10), redemption (56), or carrying value (130) of preferred stock. In the B/M sorts in June of year t, book equity is for the fiscal year ending in the preceding calendar year, t − 1, and market equity is market cap at the end of December of that calendar year. Only firms with positive book equity are used. The size premium, SMB (small minus big), is the simple average of the returns on the three small stock portfolios minus the average of the returns on the three big stock portfolios. The value premium, VMG (value minus growth), is the simple average of the returns on the two value portfolios minus the average of the returns on the two growth portfolios. VMGS is SV minus SG, VMGB is BV minus BG, and VMGS-B is VMGS minus VMGB. RM-RF is the difference between the value-weight market return and the 1-month Treasury bill rate. The table shows means, standard deviations (SD), and t-statistics for the mean (the ratio of the mean to its standard error).
| ||Factor Portfolios||Size-B/M Portfolios|
|July 1926 to December 2004, 942 Months|
| Mean||0.65||0.23||0.40||0.48||0.31|| 0.17||0.74||1.02|| 1.22||0.62||0.70||0.93|
| SD||5.47||3.36||3.58||3.63||4.25|| 3.33||7.90||7.18|| 8.32||5.47||5.85||7.35|
| t-statistic||3.64||2.06||3.43||4.08||2.23|| 1.60||2.86||4.37|| 4.49||3.49||3.69||3.89|
|July 1926 to June 1963, 444 Months|
| Mean||0.85||0.20||0.35||0.35||0.36||−0.01||1.03||1.16|| 1.37||0.83||0.89||1.20|
| SD||6.43||3.48||4.17||3.89||5.23|| 3.86||8.75||8.80||10.70||6.16||7.23||9.62|
| t-statistic||2.79||1.23||1.78||1.89||1.46||−0.08||2.47||2.78|| 2.71||2.85||2.59||2.62|
|July 1963 to December 2004, 498 Months|
| Mean||0.47||0.24||0.44||0.60||0.26|| 0.34||0.48||0.90|| 1.08||0.43||0.54||0.69|
| SD||4.45||3.26||2.96||3.39||3.12|| 2.76||7.05||5.36|| 5.39||4.77||4.26||4.44|
| t-statistic||2.36||1.68||3.34||3.97||1.87|| 2.76||1.51||3.75|| 4.47||2.03||2.82||3.49|
The size premium in average returns is similar for the two subperiods of 1926 to 2004. The average SMB return is 0.20% per month for 1926 to 1963 versus 0.24% for 1963 to 2004. It takes the power of the full 1926 to 2004 period to push the average premium (0.23%) just over the two standard error barrier (t= 2.06).
The overall value premium is also similar for the two subperiods of 1926 to 2004. The average VMG return is 0.35% per month for 1926 to 1963 and 0.44% for 1963 to 2004. The average VMG return for 1963 to 2004 is 3.34 standard errors from zero, but the average for 1926 to 1963 is only 1.78 standard errors from zero. A comparison of means test shows, however, that the premiums for the two subperiods differ by just 0.38 standard errors. Thus, there is no evidence of a change in the expected premium, and the full 1926 to 2004 period can be used to judge whether there is a value premium in expected returns. The premium for 1926 to 2004 is 0.40% per month, and it is a healthy 3.43 standard errors from zero.
Confirming Loughran (1997) and earlier evidence (Fama and French (1993), Kothari, Shanken, and Sloan (1995)), the value premium for 1963 to 2004 is larger for small stocks, 0.60% per month (t= 3.97), versus 0.26% (t= 1.87) for big stocks. But for 1926 to 1963, the value premium is near identical for small and big stocks (0.35% and 0.36% per month). Note that the difference between the small and big stock value premiums for 1963 to 2004 is mostly due to an increase in the premium for small stocks, from 0.35% to 0.60%; there is a smaller decline, from 0.36% to 0.26%, for big stocks. More important, a comparison of means test on the monthly VMGB returns shows that the big stock value premium for 1963 to 2004 is just –0.35 standard errors from the premium for 1926 to 1963, so there is little evidence of a change in the expected premium. The VMGB average return for the full 1926 to 2004 period, 0.31% per month (t= 2.23), is then solid evidence on the existence of a value premium in big stock expected returns. In short, there does seem to be a value premium in the expected returns on big stocks.
Nevertheless, the value premium in the average returns for 1926 to 2004 is 55% larger for small stocks (0.48% per month) than for big stocks (0.31%). And the average of the time series of differences between VMGS and VMGB returns is 1.60 standard errors from zero. Thus, the power of the full sample period says that there are value premiums in the expected returns on small and big stocks, but the expected premium may be larger for small stocks.
B. Finer Size Sorts
Table I classifies stocks above the NYSE median market cap as big. To facilitate comparison with Loughran (1997), we next examine value premiums for a finer size grid. We use the 25 portfolios of Fama and French (1993), formed as the intersection of independent sorts of stocks in June of each year into NYSE size and B/M quintiles. There is a problem, however. During 1926 to 1963, the portfolio for the largest size and highest B/M quintiles often has no stocks, and the portfolio for the smallest size and lowest B/M quintiles is also thin. To have at least 10 stocks in each portfolio, the tests must be limited to the period 1963 to 2004.
Table II summarizes the characteristics of the 25 size-B/M portfolios of 1963 to 2004. Specifically, the table reports monthly averages of (1) number of firms, (2) average firm size (market cap), and (3) percent of total market cap. A striking result is the skewness of percents of market cap across both size and B/M quintiles. On average the smallest size quintile contains more than half the total number of NYSE, AMEX, and Nasdaq stocks. But these smallest stocks (micro-caps) are tiny and together they account for less than 3.0% of total market cap. In contrast, there are on average just 295 stocks in the largest size quintile, but these mega-caps account for about three-quarters of total market cap. The percent of total market cap falls sharply, from 73.6% to 13.2% for the second largest size quintile and to 6.5% for the third.
Table II. Characteristics of 25 Portfolios Formed on Size and B/M: July 1963 to December 2004, 498 Months At the end of each June from 1963 to 2004, we form 25 portfolios as the intersections of independent sorts of NYSE, AMEX, and (after 1972) Nasdaq stocks into five size groups (using NYSE market cap quintile breakpoints for the end of June) and five book-to-market groups (again using NYSE quintile breakpoints for B/M). Book equity in B/M is for the fiscal year ending in the preceding calendar year and market equity is market cap at the end of December of that calendar year. Firms with negative book equity are excluded. For each portfolio, the table shows averages across the months of July 1963 to December 2004 of (1) number of firms, (2) average market cap, and (3) percent of total market cap, which is the product of (1) and (2) divided by the sum of these products across portfolios. The table also shows the average across years of B/M for each portfolio, where book equity and market equity for a given year are the sums for the firms in a portfolio. In the blocks for Number of Firms and Percent of Total Market Cap, Sum is the sum across rows or columns of the items in the column or row.
| ||Number of Firms||Average Market Cap ($Millions)|
|Small|| 532||337||338||402||645||2255|| 39|| 42|| 40|| 36|| 27|
|2|| 163||117||114||103|| 79|| 576|| 186|| 188|| 191|| 189|| 185|
|3|| 121|| 88|| 81|| 67|| 48|| 405|| 444|| 452|| 453|| 456|| 464|
|4|| 100|| 75|| 64|| 53|| 35|| 327|| 1147||1142||1150||1160||1156|
|Big|| 108|| 66|| 52|| 43|| 26|| 295||10240||7658||6608||5454||5001|
| ||Percent of Total Market Cap||Annual Sum B/Sum M|
|Small|| 0.7|| 0.5|| 0.5|| 0.5||0.7|| 2.9||0.27||0.57||0.77||1.02||1.77|
|2|| 1.1|| 0.8|| 0.8|| 0.7||0.5|| 3.8||0.27||0.54||0.76||1.00||1.66|
|3|| 1.9|| 1.4|| 1.3|| 1.1||0.8|| 6.5||0.27||0.54||0.75||1.00||1.62|
|4|| 3.9|| 2.9|| 2.6|| 2.2||1.5|| 13.2||0.27||0.55||0.75||1.02||1.62|
|Big||32.8||15.5||11.6|| 9.0||4.7|| 73.6||0.26||0.53||0.75||0.99||1.49|
Percent of market cap also falls across B/M quintiles, but not as dramatically. On average, the lowest B/M quintile (extreme growth stocks) accounts for 40.3% of total market cap, versus 8.1% for the highest B/M quintile (extreme value stocks). The decline in percent of market cap across B/M quintiles is steepest in the largest size quintile. Among these mega-caps, the lowest B/M portfolio accounts for an impressive 32.8% of total market cap, versus 4.7% for the highest B/M portfolio. This sharp decline has two sources: (1) on average the extreme growth mega-cap portfolio has about four times as many stocks as the extreme value mega-cap portfolio, and (2) though in the same size quintile, mega-cap extreme growth stocks are about twice as large as mega-cap extreme value stocks. In contrast, there is no clear relation between size and B/M in size quintiles below the largest. Except for the smallest size quintile, however, all size groups share the result that growth stocks are more numerous than value stocks. Thus, when growth and value are defined using NYSE stocks, there is a bias toward the growth in the AMEX–Nasdaq population.
For perspective on the returns we examine next, an important result in Table II is the paucity of firms that are both large and in the extreme value group. On average, only 26 firms in the size quintile of the largest firms are in the highest B/M quintile, and only 35 firms in the next largest size quintile are in the highest B/M quintile. This is not surprising, however, since firms that are large in terms of market cap are likely to have high stock prices and so are unlikely to be extreme value (high B/M) firms.
Table III shows average returns for the period 1963 to 2004 for the 25 value-weight size-B/M portfolios, along with value premiums within size quintiles, and size premiums within B/M quintiles. The value premium for a size quintile is the difference between the average return on the two highest B/M portfolios and the average return on the two lowest B/M portfolios of the size quintile. Similarly, the size premium for a B/M quintile is the difference between the average returns on the two smallest and the two biggest size portfolios of the B/M quintile. We use four portfolios (rather than the extremes of each group) to estimate premiums because, as noted above, some extreme portfolios are undiversified.
Table III. Average Monthly Returns for 25 Portfolios Formed on Size and B/M or E/P: July 1963 to December 2004, 498 Months At the end of each June from 1963 to 2004, we form 25 value-weight portfolios as the intersections of independent sorts of NYSE, AMEX, and (after 1972) Nasdaq stocks into five size groups (using NYSE market cap quintile breakpoints for the end of June) and five book-to-market or earnings-price groups (again using NYSE quintile breakpoints for B/M and E/P). Book equity in B/M and earnings in E/P are for the fiscal year ending in the preceding calendar year; M=P is the market cap at the end of December of that calendar year. (Earnings E is income before extraordinary items (Compustat data item 18) minus dividends on preferred (19) if available, plus deferred taxes from income statement (50), if available.) The size-B/M portfolios include only firms with positive book equity, and the size-E/P portfolios include only firms with positive earnings. H − L is the value premium for a size group estimated from the time series of monthly differences between the average of the returns for the two highest B/M (or E/P) quintiles within a size quintile and the average of the returns for the two lowest B/M (or E/P) quintiles. Similarly S − B is the size premium for a B/M (or E/P) quintile estimated from the time series of monthly differences between the average return for the two smallest size quintiles within a B/M (or E/P) quintile and the average of the returns for the two biggest size quintiles. t(H − L) or t(S − B) is the average monthly difference divided by its standard error. The bottom right number in the H − L columns is the time-series average (or t-statistic for the time-series average) of the overall average of the five H − L returns.
| ||Low||2||3||4||High||H − L||t(H − L)|
| Small|| 0.73||1.32||1.36||1.57||1.67||0.59||4.13|
| 2|| 0.89||1.15||1.40||1.45||1.55||0.48||3.62|
| 3|| 0.90||1.22||1.20||1.35||1.51||0.37||2.64|
| 4|| 1.01||0.99||1.22||1.34||1.37||0.36||2.75|
| Big|| 0.90||0.97||0.98||1.05||1.06||0.13||1.01|
| S − B||−0.14||0.26||0.28||0.31||0.39||0.38||3.32|
| t(S − B)||−0.77||1.46||1.85||2.18||2.63|| |
| Small|| 1.08||1.30||1.43||1.52||1.71||0.43||4.20|
| 2|| 1.07||1.31||1.34||1.36||1.53||0.26||2.00|
| 3|| 0.96||1.17||1.28||1.28||1.51||0.33||2.50|
| 4|| 0.94||1.04||1.15||1.34||1.42||0.38||3.03|
| Big|| 0.85||0.95||0.92||1.19||1.13||0.26||2.07|
| S − B|| 0.18||0.31||0.34||0.17||0.35||0.33||3.19|
| t(S − B)|| 1.05||2.04||2.36||1.33||2.54|| |
When value and growth are defined by sorts on B/M, the value premiums in average returns decline monotonically from smaller to bigger size quintiles. For the size quintile that contains the largest firms (mega-caps), the value premium for 1963 to 2004 is only 0.13% per month, and about one standard error from zero. But the value premiums of the remaining four size groups are economically and statistically substantial. They range from 0.36% to 0.59% per month and all are more than 2.6 standard errors from zero. Even in the quintile that contains the largest firms, average returns increase monotonically from lower (growth) to higher (value) B/M quintiles.
The evidence for a weak value premium in the largest size quintile depends on using B/M to identify value and growth stocks. Table III also shows value premiums within size quintiles for 25 value-weight portfolios formed on size and earnings-to-price ratios (E/P). These portfolios are formed in the same way as the 25 size-B/M portfolios, except we exclude firms with negative earnings rather than negative book equity and we use E/P rather than B/M as the value-growth indicator. The effect of this change is to tone down if not wipe out the decline in the value premium with firm size. The smallest size quintile still produces the largest value premium, but any decline in value premiums with size is far from monotonic. The largest size quintile produces the same value premium as the second smallest (0.26% per month), and the second largest size quintile produces a value premium (0.38% per month) close to that for the smallest size quintile (0.43%). Perhaps most important, when E/P is the value-growth indicator, the value premiums for all size groups are more than two standard errors from zero.
There are two interesting changes in average returns when we sort firms on E/P rather than B/M. First, and most striking, is the increase in average returns for extreme growth (low E/P) stocks in the two smallest size quintiles. This acts to reduce the value premiums for these size groups, and so brings the premiums closer to those of other size groups. In other words, the larger value premiums that we observe for small stocks when B/M is the value-growth indicator are due more to lower returns on small growth stocks than to higher returns on small value stocks. Second, using E/P as the value-growth indicator reduces average returns for the two extreme growth portfolios in the largest size quintile and increases average returns on the two extreme value portfolios. This leads to a higher value premium for the largest size quintile, and it is now more than two standard errors from zero.
Why do the return results change for the E/P sorts? Without showing the details, we can report that, at least for the smaller size quintiles, the answer traces to firms with negative earnings (excluded from the E/P sorts). Negative earnings are relatively rare for firms in the two largest size quintiles, and the average returns of large firms with negative earnings are similar to those of large firms with positive earnings. But many firms in the three smallest size quintiles are unprofitable, and their average returns are far lower than those of profitable small firms. In the two smallest size quintiles, the low returns of firms with negative earnings act mostly to lower the returns of firms in the lowest B/M quintile (extreme growth firms). This raises the estimates of value premiums for these size quintiles and creates an inverse relation between size and the value premium that is not observed when E/P is the value-growth indicator.
Finally, though not our central focus, it is interesting to note that the monotonic decline in the value premium from smaller to bigger size groups that we observe in the B/M sorts in Table III corresponds to a monotonic increase in size premiums from lower to higher B/M groups. But this pattern in size premiums for 1963 to 2004 is almost nonexistent when the value-growth indicator is E/P. Even the extreme growth (lowest positive) E/P quintile produces a hint of a size premium for 1963 to 2004, a hint that is absent in the B/M sorts.
C. International Results
International returns provide an out-of-sample test of whether there is a value premium among big stocks. Using Morgan Stanley Capital International (MSCI), we construct 25 value-weight size-B/M portfolios and 25 value-weight size-E/P portfolios using merged data for 14 markets outside the United States.1 For comparability with the U.S. results, the breakpoints for the five international size groups are the same NYSE market cap quintile breakpoints used for the United States. Since international accounting methods differ from those of the United States, the quintile breaks for B/M and E/P use the annual cross sections of the ratios for international stocks. We can report, however, that using U.S. breaks for the ratios has no effect on our inferences.
An advantage of the MSCI data is that they are free of survivor bias; firms that die remain in the historical sample. Moreover, the annual accounting data shown at the end of any month are the most recently reported, so they are publicly available. A disadvantage of MSCI is that the sample covers only 1975 to 2004. In addition, though the sample firms account for more than 80% of the market cap of the 14 markets, the small end of the size spectrum is sparse, and there are few firms in the micro-cap quintile. Thus, in presenting the international results, we show only two size groups: the top size quintile (mega-caps) and all remaining firms. This is not a problem since our main interest is whether there is a value premium for the largest stocks and whether it is smaller than for other stocks.
The international sample resembles the U.S. sample in many ways. For example, on average there are about 350 mega-cap firms in the international sample versus almost three times as many smaller firms, but the mega-caps account for about three-quarters of the sample's total market cap (Table IV). Again, more of the sample's market cap comes from growth stocks, but in the international sample, this result is due entirely to mega-caps, where growth stocks outnumber value stocks by more than two to one.
Table IV. Average Monthly Returns, Number of Firms, and Percent of Market Cap for International Portfolios Formed on Size and B/M or E/P: January 1975 to December 2004 All variables are in U.S. dollars. At the end of each December from 1974 to 2003, we form 10 value-weight portfolios as the intersection of independent sorts of international stocks into two size groups (using the market cap breakpoint between the smallest 80% and largest 20% of NYSE firms at the end of December) and five book-to-market or earnings-price groups (using international breakpoints for B/M and E/P from the combined cross section of international stocks). Book equity in B/M and earnings in E/P are for the latest reported year preceding December portfolio formation; M=P is the market cap at portfolio formation. The size-B/M portfolios include only firms with positive book equity, and the size-E/P portfolios include only firms with positive earnings. Because the smallest size quintiles have few firms, the table shows results for the largest size quintile (Big) and combined results for the four remaining size quintiles (Not Big). H − L is the value premium for a size group estimated from the time series of monthly differences between the average of the returns for the two highest B/M (or E/P) quintiles within a size group and the average of the returns for the two lowest B/M (or E/P) quintiles. t(H − L) or t(Big − Not Big) is the average monthly difference divided by its standard error. Percent of Market Cap is the monthly average of the percent of total sample market cap in each portfolio. The 14 international markets in the tests are Australia, Belgium, Canada, France, Germany, Great Britain, Hong Kong, Italy, Japan, the Netherlands, Singapore, Spain, Sweden, and Switzerland.
| ||Average Monthly Returns|
|All||Low||2||3||4||High||H − L||t(H − L)|
| All|| 1.11|| 0.84|| 1.13|| 1.23|| 1.40|| 1.63||0.53||2.63|
| Big|| 1.07|| 0.81|| 1.14|| 1.21|| 1.41|| 1.56||0.51||2.24|
| Not Big|| 1.26|| 1.00|| 1.15|| 1.26|| 1.34|| 1.69||0.44||2.66|
| Big – Not Big||−0.19||−0.18||−0.01||−0.05|| 0.07||−0.13||0.06||0.41|
| t(Big – Not Big)||−1.49||−1.13||−0.10||−0.40|| 0.52||−0.75|| |
| All|| 1.12|| 0.72|| 1.08|| 1.23|| 1.45|| 1.65||0.65||2.78|
| Big|| 1.08|| 0.68|| 1.09|| 1.17|| 1.44|| 1.57||0.62||2.50|
| Not Big|| 1.29|| 0.86|| 1.15|| 1.34|| 1.45|| 1.77||0.61||3.02|
| Big − Not Big||−0.21||−0.18||−0.06||−0.18||−0.01||−0.21||0.01||0.11|
| t(Big − Not Big)||−1.64||−1.09||−0.44||−1.38||−0.07||−1.42|| |
| ||Percent of Market Cap||Average Number of Firms|
| All||100.0||33.7||25.8||19.2||13.8|| 7.5||1421||289||289||287||282||274|
| Big|| 73.4||27.8||19.9||13.4|| 8.5|| 3.6|| 362||115||100|| 72|| 49|| 25|
| Not Big|| 26.6|| 5.9|| 5.8|| 5.8|| 5.2|| 3.9||1059||174||188||214||233||249|
| Big|| 74.5||23.2||17.6||14.0||11.7|| 8.1|| 341|| 96|| 82|| 66|| 57|| 41|
| Not Big|| 25.5|| 5.4|| 5.5|| 5.3|| 5.1|| 4.3|| 906||163||175||186||192||192|
More important, Table IV documents strong value premiums in international returns. (As in the U.S. results, we estimate international value premiums as the difference between the average returns for the two extreme value and the two extreme growth portfolios of a group.) When B/M is the value-growth indicator, the overall value-weight international value premium is 0.53% per month (t= 2.63); it is 0.65% per month (t= 2.78) for E/P groupings. (See also Fama and French (1998).)
The new evidence in Table IV centers on the value premium for very large stocks. When we sort on B/M, the value premium for mega-caps is six basis points per month larger than it is for all smaller stocks, but the difference is indistinguishable from zero (t= 0.41). Sorting on E/P, the premiums for the two size groups are virtually identical. In short, international returns show economically and statistically strong value premiums, and they are as large among the biggest stocks as among smaller stocks.
D. Bottom Line
In sum, when we use B/M to identify value and growth stocks in the United States, value premiums for 1963 to 2004 are smaller for big firms. Although there are large and statistically reliable value premiums in the four smaller size quintiles, the premium for the largest size quintile, 0.13% per month, is just 1.01 standard errors from zero. When we sort on E/P rather than B/M, however, we find a strong value premium in the largest size quintile and little relation between the value premium and firm size. The evidence for 1926 to 1963 (Table I) is also relevant. Though we only have a 2 × 3 size-B/M grid for the earlier period, if the value premium is negatively related to size, small stocks should produce a bigger premium than big stocks. But the value premiums for small and big stocks for 1926 to 1963 (0.35% and 0.36% per month) are near identical. Finally, whether we sort on B/M or E/P, international average returns for 1975 to 2004 show strong value premiums that are at least as large for mega-caps as for smaller stocks. All this suggests that the weak relation between B/M and average returns observed for the largest U.S. size quintile for 1963 to 2004 may be a random aberration, due perhaps to the fact that there are relatively few mega-cap value stocks.