What is the effect of adding a dummy variable into the logit model, where the dummy ==1 if currently married, and ==0 if not? Do any of the other coefficients change? How can you interpret this coefficient? Is it statistically different from zero?
To get the variables in order, we need to generate a dummy
variable for whether a person is married or not. Let's do this by first
investigating the variable for marital status:
tab marstat tab marstat, nol /*we need to know how the variable is coded to create the dummy variable*/ gen married=0 if marstat!=. /*if marstat==., we want married==. too*/ replace married=1 if marstat==1 lab val married yesno /*now we can attach a value label to this dummy variable*/
Now our new variable is in order, we can re-run the same logit regression, with this dummy variable included:
char literacy[omit] 3 char location[omit] 3 xi: logit part1con1 age female i.location femeduc educ i.literacy married
And the logit output is then:
Logit estimates Number of obs = 1738
LR chi2(9) = 713.94
Prob > chi2 = 0.0000
Log likelihood = -781.69863 Pseudo R2 = 0.3135
------------------------------------------------------------------------------
part1con1 | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
age | -.098227 .0077152 -12.73 0.000 -.1133486 -.0831055
female | -1.229553 .3618651 -3.40 0.001 -1.938795 -.5203099
_Ilocation_1 | .2720511 .1500643 1.81 0.070 -.0220697 .5661718
_Ilocation_2 | .2579904 .1647732 1.57 0.117 -.064959 .5809399
femeduc | .086098 .0375943 2.29 0.022 .0124145 .1597814
educ | .0726885 .0267122 2.72 0.007 .0203335 .1250434
_Iliteracy_1 | 1.239322 .4662134 2.66 0.008 .3255602 2.153083
_Iliteracy_2 | .7513223 .4727324 1.59 0.112 -.1752163 1.677861
married | -1.611888 .1669915 -9.65 0.000 -1.939185 -1.284591
_cons | 2.390094 .5154936 4.64 0.000 1.379745 3.400443
------------------------------------------------------------------------------
All of the coefficients do change size, when we include the married dummy, but none of them change sign. Some coefficients become significant, whereas they were not in the original regression.
The coefficient on married is indeed significant at the 1% level. It is negative, implying that married individuals are less likely to report that they used a condom the first time they had sex with their most recent partner. For many married people, the most recent partner is probably the spouse. If we use the formula: marginal effect = F(Xb)*(1-F(Xb))*b where F(Xb) = the sample mean for the variable part1con1, we get the following:
summ part1con1 /*this will provide the sample mean that we need*/
di "the mfx of being married on the prob of using a condom the first time with the most recent partner is " .5754625*(1-.5754625)*-1.611888
And the output is:
the mfx of being married on the prob of using a condom the first time with the most recent partner is -.39379296
So the answer is -.39379298. There is a 40% reduction in your chances of reporting YES to the question if you are married! How can we think about what this implies about the level of trust in marital relationships?