I would like to check if there is a relationship between mothers' fertility and grandparental childcare. I am using a cross-national survey dataset. I restricted the sample to 6 countries of interest and only kept female respondents between fertile ages, who live with a partner and who had at least one child under 14 years old at time t.

Dependent variable: dummy taking value 1 if the respondent had an additional birth between time t and t+1

Main explanatory variable: dummy taking value 1 if the respondent received help with childcare from her parents at time t

Control variables for respondent, all measured at time t: - age (numeric variable, ranging between 18 to 45) - education (two dummy variables: first takes value 1 if highest completed schooling is upper secondary; second takes value 1 if highest completed schooling is tertiary; this means the reference level is if highest completed schooling is low secondary) - employment status (dummy taking value 1 if respondent is employed) - financial stability (dummy taking value 1 if respondent reports getting by financially with ease) - conservative views (dummy taking value 1 if respondent agrees with a given statement) - number of biological children (numeric, between 1 and 3) - age of youngest child (numeric, between 1 and 14)

I intend to use a logit model in Stata and I would like to report separate coefficients for each country.

The nr of observations per country range from a minimum of n=180 to a maximum of n=1100. The proportion of respondents who exhibit the outcome of interest per country ranges from a minimum of 6.5% to a maximum of 12% (for one of the countries, this means only 15 cases had an additional birth)

1. Are these enough observations per country to run a logit model and obtain meaningful estimates?

2. If the answer to 1. is positive, would it be correct to then run the following command and interpret the resulting coefficients: by country, sort: logit additionalchild grandparenthelp age upsec_edu tertiary_edu employed fin_stable conservative nrbiokids ageyoungest

In other words, does running the regressions by country solve possible issues with clustering?

3. If the answer to 1. is negative, and I instead run a unique regression (command as above but without the by country prefix), then I would have to use the option vce(cluster country), is this correct?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.