a regression coefficient k, then again the Pareto lav. 1 of 
the wealth distribution will be reproduced in incone, 
but this tine the Pareto coefficient will be modified to 
<*/k. 
If k is below unity - which we may anticipate, is in 
reality the likely case - then the Pareto coefficient for 
income will be larger than for wealth. 
It is time now to turn to the restrictive assumptions which] 
so far have been stated only in algebraic terms: 
\Oy in the case of independence, kV^ Y in the case of 
linear dependence. 
,C, 
This means that the rate of return must not be 100 j# or 
t 
larger in the first case; in the second case, if k<1, 
■n-ijn [. i'j. ..7.Mi4-3A%T ftiri ; 
wealth is defined as equal to zero for ¥<0; in con 
sequence, the left tail of the function f* (W) (correspond 
ing to negative values of W, thus to rates of return of 
\ 
100 % and more) must also be defined as equal to zero (see 
fig. 2). The function f* (W) relates to the case where 
Hv s t'j i ) * 
1