In: Statistics and Probability
a) Usually in statistics as well as in econometrics we use sample values to infer population parameters.
So for example in linear regression we don't know what are the actual population parameters of regression equation so we try to estimate the population regression line( parameters) using some data/sample. Once we have the estimates it becomes necessary to test the parameters if they are significant or not. Specially in multiple regression where there are more variables it becomes necssary to include only those which are significant so that is tested by hypothesis.
In statistics we usually conduct hypotheses tests to check if population means assumes a specific value or not?
whether difference of proportion is significant or not?
and much more .
So basic thing is that in both statistics and econometrics we rely heavily on sample data and we need to infer the population parameters usings sample data so to do that with confidence we need these statistical tests.
Now error term is included in the model to account for random fluctuations that arise because of several reasons.
for example your model might not include all the variables . In simpler language our model is simplification of reality and not all variables are included so the error term accounts for their average effect.
Also there might be some error due to collection of data so basically error term shows the average effect of these things.
I hope it helps you
Thanks