ECO 206-02
Spring 2017 Ashoka University
Take-home STATA Practical SET A
7th May, 2017
1
1. A random sample of 753 women were selected and surveyed (MROZ.dta). We are interested in what
determines the womans choice between working and staying at home (which for our purposes is classi
ed as not working). The following variables were collected:
(a) Estimate a Linear Probability model using the last 7 variables and a constant to explain the labor
force decision of the women. Call this model LPM_Full
(b) For each of the following variables indicate whether the variable has a positive or negative eect
on a womans labor force participation and test whether the eect is dierent from zero at the
5% level of signicance (normal tables are attached as well).
1. Kids less than 6
2. Education
3. Non wife income
(c) Check if experience has a non-linear eect of women labor force decision.
(d) Suppose we conjecture that age, education and experience has no bearing on women labor force
decision. Call this model LPM_partial. Check if these restrictions are valid. Clearly state the
Null, the test statistic, the conclusion.
(e) Estimate a third model assuming that all variables except the constant have a zero coe cient.
Call this LPM_Constant. Report the results from three models in the standard format including
signicance of each variable, and standard errors.
(f) Using this last model, derive an expression for the probability that inlf = 1 as a function of the
coe cient on the constant and use the expression to compute the proportion of working women
in the sample. What does this imply about the number of working women in the sample of 753
women?
(g) Perform joint tests of signicance of the variables (except the constant) in the rst two models.
2
. describe
Figure 1: CRIME1.DTA
2. Consider the dataser CRIME1.DTA to analyze the factors that may lead to an arrest. The dataset is
decribed as follows ( Figure 1):
(a) Create a variable arr86 which is an indicator of whether a male was arrested at least once in 1986.
Consider the linear probability model for arr86. Call this Model 1. Present the results in standard
format.