1.     You have two data sets:  one from Iowa, and one from Nebraska.  Each data set began in 1960, with data collected on 100 family farms.  Subsequent data was collected in five-year intervals in 1965, 1970, 1975, 1980, 1985, 1990, 1995, 2000 and 2005.  Variables included:

farm size  crops grown  gov't subsidies

organic techniques used? storm damage to crops annual rainfall

farm income fertilizers used days of sunshine

crop yield average temperature labor costs

 

The Iowa data is panel data (that is, you have data on the same, identifiable farms for the 8 periods).  The Nebraska data is not a panel (each year, a random sample of 100 farms was drawn). You are interested in the factors that influence farm failure and bankruptcy.   What types of analysis can you do with the Iowa data that you cannot do with the Nebraska data?  Conversely, what problems might you have in analyzing the Iowa data, especially in later years?

 

2.     Suppose you examined block-level data for Gotham City on the links between crime, income and race, and you created the following correlation matrix.

	Crime Rate	Percent Minority	Income
Crime Rate	1.00	0.42	-0.54
Percent Minority Population		1.00	-0.74
Income			1.00

And you also calculated the following partial correlation:

Variables                                                                                     Controlling for.....                           correlation

crime and minority                                                                            income                                             -0.13

How do you interpret this data?  What is the relationship here between race and crime?  What assumptions do you make about causality? 

 

3.     A researcher has two data sets on the relationship between housing prices and distance from the central business district (CBD) in the 50 largest American metropolitan areas.  The first data set's unit of analysis is the individual house (n=1000), and shows a positive relationship between house price and distance to CBD.  The second data set's unit of analysis is the city (n=50), and shows a negative relationship between the city's mean housing price and the mean distance of housing from the CBD.   Why are the two results different?  Which set of results is more useful?  What danger does a researcher face in generalizing from the second data set? 

 

4.     The Census Department releases the results of a  study on how women of four different ethnic groups fare on unemployment and wages in 25 American cities from 1990 to 2006.  The results show that all four groups do better when  1)  Overall employment growth is relatively high, and 2) the share of each group in the labor force is relatively small.  All other explanations of unemployment and wage differences turn out to be statistically insignificant.    

The report's policy section states:   "Minority women would be better off if the government stimulated growth rather than pursued affirmative action programs or other race-specific remedies.  In addition, minority women should attempt to move to cities where the ethic minority shares are relatively small compared to the size of the overall labor force."        Comment on the structure of the research design and on the policy conclusions.