### Table 6: Mean regression decompositions of the race gap in wealth (siblings sample) White coefficients Black coefficients

"... In PAGE 30: ... The comparison of the estimates with and without fixed effects suggests that parental wealth or other family background factors can explain about one third of the huge race gap in the relationship between self employment and wealth.29 In Table6 we report decompositions of the wealth gap using the siblings sample using FE and FE-Means estimates of the wealth models. For purposes of comparison, we also report OLS and OLS-Means estimates.... ..."

### Table 8: Mean regression decompositions of the race gap in growth of wealth (5 year changes, couples sample)

"... In PAGE 32: ... We do not weight observations in computing the decompositions. The wealth change decompositions are reported in Table8 . In the case of W , in- come and demographic variables explain 74% of the difference in accumulation rates when the accumulation model for whites is used.... In PAGE 32: ... This fact, in combination with the substantial race differences in the wealth growth models, suggests that differences in savings behavior (possibly in combination with rates of return) may be an important source of the wealth gap and of black/white differences in the wealth models.32 Note that the results for logs in Panel C of Table8 present a challenge to the above interpretation. The results for logs show that 77% of the wealth is explained when the black wealth growth model is used while only 5% of the gap is explained when the white log wealth growth model is used.... ..."

### Table 4: Median regression decompositions of the race gap in wealth White coefficients Black coefficients

"... In PAGE 20: ... We refer to this as the gap in the conditional median of wealth. Table4 presents the decompositions. In the case of couples, the total gap in the median is estimated to be 84,794 (2,824), which compares to a gap in the mean of wealth of 116,795 (4,535).... In PAGE 21: ... When the median wealth model for blacks is used, the percentage explained is slightly lower than the corresponding value for the conditional mean in the case of couples but higher for single men and women. The results for W=y in Panel B of Table4 are similar to the results for levels, except that the portion of the gap explained by the white median regression model is substantially smaller for both couples and single females. The results for ln W in Panel C of Table 4 are basically similar to the results based on mean regression and also imply that the wealth of whites is much more sensitive to income and demographics than the wealth of blacks once the implications of the log specification and the large unexplained gap are taken into account.... In PAGE 21: ... The results for W=y in Panel B of Table 4 are similar to the results for levels, except that the portion of the gap explained by the white median regression model is substantially smaller for both couples and single females. The results for ln W in Panel C of Table4 are basically similar to the results based on mean regression and also imply that the wealth of whites is much more sensitive to income and demographics than the wealth of blacks once the implications of the log specification and the large unexplained gap are taken into account. 8 Self Employment and the Wealth Gap If causality runs mainly from self employment to wealth, then it is desirable to control for self employment in the analysis, especially to the extent that the effects of past discrimination on the self-employment rate of blacks lingers today (see for example the discussion in Bates (1997) and Oliver amp; Shapiro (1997)).... ..."

### Table 5: Mean regression decompositions of the race gap in wealth components (couples sample)

"... In PAGE 23: ... (The model estimates are based on the full samples.) The results are presented in Table5 of Altonji amp; Doraszelski (2001). They confirm our basic finding that differences in the white and black wealth models play a key role in the wealth gap.... In PAGE 25: ... 10 Race Differences in the Components of Wealth Why are the coefficients so different across the black and white specifications? One hypothesis is that the relationship differs for some forms of wealth but not for others. Table5 reports mean regression decompositions of main home equity (house value net of mortgage balance), wealth in farms/businesses, and stocks/mutual funds and IRAs. We restrict attention to couples.... In PAGE 25: ...RAs. We restrict attention to couples. The regressors are the same as the ones used for couples in our other wealth models. On average whites hold 32,309 (978) more home equity than blacks (56,859 (657) versus 24,550 (724), Panel A of Table5 ), which is not surprising given the difference in home ownership rates across races.23 The white regression model explains 88% of this gap compared to the 79% explained for total wealth in Table 3.... ..."

### Table 3: Varying the size of the time gap between timeslots. 8 timeslots per frame. Time gap

"... In PAGE 8: ... In conclusion, the smaller the demands are, the larger the gain is of using SOH. In Table3 we find the wavelength usage as function of the gap when the number of timeslots is 8. It can be seen that the wavelength usage is approximately inversely proportional to 1BP8 A0 g, where g is the length of the gap.... ..."

### Table 4: Failure rates for hierarchical alignment of the MTEval bitext, over varying numbers of gaps and constrain- ing trees (CTs).

"... In PAGE 5: ... The absence of such cases in the data implies that the failure rates under the constraints of one parse tree would be identical even if we allowed production rules of rank higher than two. Table4 shows the alignment failure rates for the MTEval bitext. With word alignment constraints only, 3% of the sentence pairs could not be hierar- chically aligned without gaps.... ..."

Cited by 1

### Table 4: Failure rates for hierarchical alignment of the MTEval bitext, over varying numbers of gaps and constrain- ing trees (CTs).

"... In PAGE 5: ... The absence of such cases in the data implies that the failure rates under the constraints of one parse tree would be identical even if we allowed production rules of rank higher than two. Table4 shows the alignment failure rates for the MTEval bitext. With word alignment constraints only, 3% of the sentence pairs could not be hierar- chically aligned without gaps.... ..."

Cited by 1

### Table IV. Evolution of the achieved wealths according to the tuning parameter of eg and b1exp both for fixed and time varying parameters. Computations are realized on random samples of size 100, arithmetic means are displayed. Monthly rebalancing.

2005

Cited by 9

### Table V. Evolution of the achieved wealths according to the tuning parameter of eg and b1exp both for fixed and time varying parameters. Computations are realized on random samples of size 100, arithmetic means are displayed. Daily rebalancing.

2005

Cited by 9

### Table IV. Evolution of the achieved wealths according to the tuning parameter of eg and b1exp both for fixed and time varying parameters. Computations are realized on random samples of size 100, arithmetic means are displayed. Monthly rebalancing.

2005

Cited by 9