Why are some firms more profitable than others?

Find data to help answer this question. I want you to combine firm-level data from Compustat with data on industries or time periods, to give you practice in creating a dataset. Thus, please also test for whether the following matter to a firm's profit: (a) GDP growth, and (b) the distribution of employment per enterprise in the industry or subindustry (for that data, see http://www.census.gov/csd/susb/susb02.htm). You do not need to use a structural model.

You may do this in groups of two or three, turning in one assignment. Different groups may consult with each other.

Some steps you should take:

1. Define profit.

2. Decide which firms will be in your sample (one industry? many industries? different years? all sizes?)

3. Decide what variables affect profit, and how.

4. Do statistical tests of some kind.

5. Check to see if your model is robust by doing such things as trying nonlinear specifications, seeing the effect of outlier observations, and splitting the sample.

6. Write up your results. Provide summary statistics of some kind, and at least one diagram. Do not simply reproduce regression program output; think carefully about what information to present to your reader. Interpret the size of coefficients as well as their significance.

You will not come up with a publishable paper as a result of this exercise. Rather, I want you to practice addressing a question with data. It is fine just to examine how just one variable affects profit (besides the two I specified above).