1. Suppose you are interested in researching the relationship between militaryservice and criminal behavior. Sociological studies contend that military experiences have varying effects on different people. For some, the military is an opportunity for career advancement, whereas for some others it may have an adverse impact on their behavior that results in increased aggression.
   1. Discuss what measures you might use to quantify criminal behavior and military experience. What kind of data would you need (hint: think about control variables, control groups, period of observation, panel/pooled/cross-sectional data)?
   2. Propose a model that will allow you to estimate the effect of military service on criminal behavior.
   3. Modify the above model to allow the effect of military service on criminal behavior to vary by gender and race. You should also allow gender and race to affect criminal behavior on their own.
      
   4. Do you anticipate any selection problems or potential bias from unobservables? Describe what they would be and what may be some solutions.
2. Use data in WAGE1.DTA for this exercise.

1. Estimate the following model:
   \[\log (wage)={{\beta }_{0}}+{{\beta }_{1}}female+{{\beta }_{2}}educ+{{\beta }_{3}}exper+{{\beta }_{4}}expe{{r}^{2}}+{{\beta }_{5}}tenure+{{\beta }_{6}}tenur{{e}^{2}}+u\]
   What is the interpretation of \[{{\beta }_{1}}\] ?
2. Perform a Ramsey RESET test on this model, report and interpret the result.
3. Some researchers contend that women earn less because they accumulate less firm-specific human capital (less tenure). Does the data indicate that women have less tenure than men? Is this difference significant at the 5% level?
4. Test whether the return to specific human capital is lower for females compared to males (only interact female with the appropriate variable in level form, not the squared term).
5. What is the interpretation of the coefficient of female in the model estimated in (d)?
6. Use the model estimated in (d) to calculate the gender differential when tenure = 5.10 years. Is this differential statistically significant?
7. Now estimate the model:
   \[\log (wage)={{\beta }_{0}}+{{\beta }_{1}}female+{{\beta }_{2}}nonwhite+{{\beta }_{3}}educ+{{\beta }_{4}}exper+{{\beta }_{5}}expe{{r}^{2}}+{{\beta }_{6}}tenure+{{\beta }_{7}}tenur{{e}^{2}}+u\]
   Do nonwhites seem to earn significantly different wages than whites?
8. Test whether white females earn significantly lower wages than white males.
9. Test whether white females earn significantly different wages from nonwhite females (Hint: you may have to run a different regression).

3. Use the state-level data on murder rates and executions in MURDER.DTA for the following exercise.

1. Consider the following model
   \[mrdrt{{e}_{it}}={{\beta }_{0}}+{{\delta }_{0}}d90+{{\delta }_{1}}d93+{{\beta }_{1}}exe{{c}_{it}}+{{\beta }_{2}}unem+{{u}_{it}}\]
   Discuss what the signs of the coefficients are expected to be.
2. Estimate the model using OLS ignoring the longitudinal nature of the data. Do you find executions to have a deterrent effect?
3. Test for heteroskedasticity in the errors of this model. Use the White test, and interpret the results.
4. Using the model in (i) test whether the effect of unemployment on murder rates has changed over time.
5. Using the model in (i) but excluding year 1990 (so keep only observations from 1987 and 1993) check whether there have been any structural changes in the relationship between violent crimes and executions between 1987 and 1993.

Price: $31.18

Solution: The downloadable solution consists of 15 pages, 1618 words and 13 charts.
Deliverable: Word Document