Log-lin model in simple regression: yi=B0.B1Xi.eui   take natural log of both sides lnyi=lnB0 + (lnB1)xi +…

Log-lin model in simple regression:

yi=B0.B1Xi.eui  
take natural log of both sides
lnyi=lnB0 + (lnB1)xi +
ui . Take lnyi = wi,
lnB0= B*0 and
lnB1=B*1 so the former equation
equals to wi= B*0 +
B*1 + ui. It is said that, when
you increase x by 1 unit, you expect B*1 %
change in y.
B*1
=(dyi/yi​) / dxi =
= (((yi-yi-1)/yi) /
(xi- xi-1))*100 . question is about the
numerator, (yi-yi-1)/yi)*100:
1)It is said that the numerator reflects the percentage change
in y, then why it is not
((yi-yi-1)/yi-1) but instead
((yi-yi-1)/yi) ? Why don’t we
divide by yi-1 (the value before i-th term,or the past
value) instead of yi (the i-th term, or the current
value) ?
2) Does ((yi-yi-1)/yi)*100
really stand for the percentage change in y?