s*****n
发帖数: 839 | 1 I'm doing a research on predicting unemployment rate by a causal model. I
have my original ideas of how to do it. But the ideas are just the baby
plant now. Need more input to drive them into a good model.
If anyone interested, please give me some advice on how to get the model
done from scratch.
I'm trying to predict US unemployment by national, state and county in short
run and in long run.
I think if we want to find a causal model which is good, then we need to
figure out what lead to unemployment.
My ideas are: unemployment come from two major reasons:
1. The market don't need so many labors in certain location no matter how
good you are.
2. The market need people with certain skills and experience but such
candidates are not found in US labor force.
So the number of business and the size of the company and overall economy
will determine the first one and people's education and work experience and
industry will determine the second one.
I have this idea but don't really know how to get it carried out to build a
sound model. From BLS I can find unemployment rates. With our own mapping
algorithm we can get the numbers from state or county to zip. With market
index we can model the economy. This is not the problem.
The problem come from the data involving business number and their average
size by location, how to get the education data and how to build it into the
model, how to get experience data and how to get it into model. And plus,
we need the granularity of location and industry.
I think we can first model by industry and sum them up by location.
Anyone is working on similar research or job has an idea or input?
Thank you very much. |
|
