### Table 2 Maximum Likelihood Estimates of Mixed Gamma Distribution on the Excess Market Returns Entire Sample Period Sub-period 1 Sub-period 2

### Table 13 Comparison of passive and active investment strategies in the maximally predictable port- folio for SBU, SIZE, and SECTOR asset groups, for monthly returns over the sample period from 1967:3 to 1990:12. WT reports the total return over the entire sample period, and the breakeven cost is the one-way transaction cost which equates the total returns of the active and passive strategies.

1991

"... In PAGE 31: ...ections 5.1 and 5.2. Table13 shows that the active asset allocation strategy outperforms the passive, yielding a higher mean return, a lower standard deviation of return, and a larger total return WT over the investment period. For example, a $1 passive investment in the SIZE MPP grows to $16.... In PAGE 31: ...trategy yields a remarkable $103.52. Of course, the values of WT for the active strategy do not include transactions costs, which can be substantial. To determine the importance of such costs, Table13 also reports a breakeven cost quot; defined to be that percentage cost s of buying or selling the MPP which would equate the active strategy apos;s total return to the passive strategy apos;s. More formally, if the active strategy calls for k switches over the 279-month investment period, then the one-way breakeven transaction cost s is given by: - 29 - 13.... In PAGE 32: ... (1 -s )k ) i/k (5.10) Table13 shows that the one-way transaction cost would have to be somewhere between 1.21 percent and 3.... In PAGE 32: ....21 percent and 3.13 percent for the active strategy to yield the same total return as the passive. We cannot conclude from Table13 that the MPP is a quot;market inefficiency quot; which is exploitable by the average investor, since we have not formally quantified the (dynamic) risks of the passive and active strategies. Although the active strategy apos;s return has a lower standard deviation and a higher mean, this need not imply that every risk-averse investor would prefer it to the passive strategy.... ..."

### Table 1 Sample summary statistics and the distribution of block trading volume The Paris Bourse sample consists of the component firms of the SBF-250 index that trade common stock in the continuous auction market on April 1, 1997. The sample period is from April 1997 to March 1998 and the data source is the Base de Donnees de Marche (BDM) database. The firms are classified into activity quintiles based on normal block size (NBS), a measure of the trading activity in the stock during the sample period. Only block trades executed during regular market hours are included in the analysis. Reported are the average market price and market capitalization of the sample of firms on April 1, 1997. For upstairs and downstairs block trades, the table reports the total number of trades, the mean and median trade size, the cumulative trading volume, and the percentage of the trades and cumulative trading volume executed in the upstairs market during the entire sample period.

2002

"... In PAGE 13: ...3. Descriptive statistic on Paris block trading Table1 presents sample summary statistics. Sample firms are classified into activity quintiles.... In PAGE 13: ... Average market capitalization increases monotonically from FF 1,614 million for the least active quintile to FF 48,670 million for the most active quintile. (Insert Table1 near here) The sample includes 92,170 block trades. Of these, 31,088 (33.... ..."

### Table 14. Economic quantities over entire period.

in OF

"... In PAGE 7: ...able 13. Sample economic data......................................41 Table14 .... ..."

### Table 2 shows the sum of the squared errors and R2 for the entire 120-quarter period, the 40-

1995

"... In PAGE 19: ...Table 2 shows the sum of the squared errors and R2 for the entire 120-quarter period, the 40- quarter out-of-sample period, and the 80-quarter in-sample period. Table2 . The sum of the squared errors and R2 for the 120-quarter period, the 40-quarter out-of- sample period, and the 80-quarter in-sample period.... ..."

Cited by 3

### Table 4: Periodic Correlations one step ahead forecasts errors independent PUC Model (22)

2006

"... In PAGE 19: ... Therefore we reestimate the model for the entire sample and look at the residual periodic autocorrelations. Table4 presents the periodic autocorrelations for the one-step-ahead forecast errors of 1995.... ..."

Cited by 2

### Table 4: Periodic Correlations one step ahead forecasts errors independent PUC Model (22)

2006

"... In PAGE 21: ... Therefore we reestimate the model for the entire sample and look at the residual periodic autocorrelations. Table4 presents the periodic autocorrelations for the one-step-ahead forecast errors of 1995.... ..."

Cited by 2

### Table 8: Summary Statistics. Jan. 27 1984 to Dec. 31, 1996. JPM MMM MCD INTC IBM XRX XON

"... In PAGE 18: ...f the underlying asset. This extends to the tails of returns. In Table 12 we present summary statistics on a wide range of financial returns for the period 1987-1996, and is clear that the tails all have similar properties. Summary statistics for each stock return are listed in Table8 for the entire sample period, and in Table 9 for the 1990- 1996 testing period. The corresponding correlation matrixes are presented in Tables 10... ..."

### Table 12: 10 Years, 2600 Daily Returns, 1987 - 1996. With Predicted Maximum Daily Drop in one Year (250 days)

"... In PAGE 18: ... This extends to the tails of returns. In Table12 we present summary statistics on a wide range of financial returns for the period 1987-1996, and is clear that the tails all have similar properties. Summary statistics for each stock return are listed in Table 8 for the entire sample period, and in Table 9 for the 1990- 1996 testing period.... ..."

### Table 3 reports the percentage of exceptions observed for each of the 24 VaR models for the 1%, 5%, 10% and 25% VaR estimates over the entire out-of-sample period of 1607 observations. These summary statistics are key components of the LRuc and LRcc test results reported in Table 4. Both the tables and the discussion below are framed with respect to the distributional assumption first and then with respect to the relative performance of the six covariance matrix forecasts and two portfolio variance forecasts. As shown in Panel A of Table 4, VaR models based on the standard normal distributional assumption perform relatively well at the lower quantiles (1% and 5%) in that only a few forecasts fail the LRuc and LRcc tests at the 5% significance level. However, almost all fail at the higher quantiles (10% and 25%). As shown in Panel B, VaR models based on the t(13) distributional assumption perform well only for the 1% VaR estimates. This result indicates that

2000

"... In PAGE 34: ... Table3 . Observed Exception Rates for the 24 VaR Models Observed frequency of exceptions Panel A.... ..."

Cited by 1