If the sample size is smaller enough to make make the population size infinite i.e. 20 times larger then the sample size then the <u><em>co-variance</em></u> ( a property of retaining the original form of a function when variables are transformed linearly) <u><em>becomes nearly equal to zero</em></u>. In this way sampling with replacement (large population) is not different from the sampling without replacement (small and finite population) as co-variance is close to zero. <em><u>Thus they are nearly equal</u></em>.
For example:
The simple sampling with replacement is that every individual has the same probability of being chosen i.e. for a small sample from a large population, while sampling without replacement is approximately the same, since the chances of choosing the same individual is low.