In my research methods course project, I'm investigating the impact of attending a private school during primary and secondary education on future wages, utilizing cross-sectional data. I've encountered a situation where some individuals in my sample are college students who do not currently earn a wage. How should I address this issue? should i exclude them from the analysis?
2
-
1$\begingroup$ "impact ... on future wages". Are you considering total lifetime wages, or just wages at one time, eg in a particular year? $\endgroup$– Adam BaileyCommented Jan 15 at 16:36
-
$\begingroup$ wages at one time $\endgroup$– baker MCommented Jan 15 at 16:49
Add a comment
|
3 Answers
$\begingroup$
$\endgroup$
1
The best approach here will depend on the contents of your sample.
If the sample includes individuals of all ages, or at least a good range of ages, then my suggestion would be to exclude all individuals whose age is below that at which most people who attend college have completed their studies and joined the labour market. The alternative of excluding only those attending college may result in bias because for individuals to attend college may not be independent of the type of school they attended. Many people progress to higher income jobs over the course of their careers, so those who are working at an age when many others are at college will tend to have lower incomes than older workers. If, for example, those who attended private school are more likely to attend college, excluding individuals at college but not others of a similar age would therefore tend to raise the average incomes within the sample of those who attended private school relative to others.
If on the other hand the whole sample consists of individuals at 'college age', then there is little alternative but, as csilvia says, to exclude those at college. Whether the results would be meaningful would be questionable, especially if a high proportion of individuals attended college.
-
$\begingroup$ Great point! Thank you for sharing! $\endgroup$ Commented Jan 23 at 18:47
$\begingroup$
$\endgroup$
If you only have cross-sectional data and no data on wages of these students then it is sensible to exclude them from your dataset.
Usually it is not recommended to drop observations from data as you can be accused of cherry-picking or data mining, but for these observations the dependent variable simply does not exist.
Alternative approach would be to estimate the potential wage these students could get if they would be employed, but it is debatable whether this would significantly improve the research. Some people could even argue its worse than deleting the observations, and estimating the missing data would take a lot of work. Consequently, I would recommend just dropping the data.
$\begingroup$
$\endgroup$
In my country, Medical Doctors lifetime earnings pass that of Plumbers when the two groups reach their late 40's. Wages converge sometime earlier, but it takes time to catch up on lost earnings. School teachers never catch up with plumbers.
If you compare the point-in-time wages of Plumbers and Medical Students, the effect is real, and has a real effect on life time earnings, and is a real effect of private-school education: at age 21, private school students earn less than trade school graduates.
The problem you are observing is a general problem of point-in-time statistics, and has no universal solution. You get different answers to different questions. You should do both calculations, and, possibly, the numbers will be similar for both cases. If you're lucky, the answer will be dramatically different for the two calculations, and you'll have an interesting observation to add to your report.
