Thursday, May 04, 2006

Everybody's Doin' It: Tutorial Blog

Everybody's Doin' It: Tutorial Blog

Every year, as we come closer to the end of the term, students at Harvard worry about finding an impressive summer internship. This process comes with a lot of costs including time to find applicable internship positions, applying for these positions, which is time consuming, interviewing, which is scary and stressful, and preparing for the job by learning about the company and the tasks that lie ahead, which all clearly involves time and effort. Why do students go through all of this? Most, especially at competitive schools, believe that holding these summer internship positions will make them much better candidates for work following graduation, and will amount to a significantly higher income in the future. It appears though that at less selective colleges, students are much less likely to incur the costs of these positions. First, I want to test how strongly school selectivity determines whether a student will hold a summer internship. Then, I want to observe if in actuality the benefits of obtaining a summer internship outweigh the costs. The way I will observe this is by looking at both success in school following a summer internship and success in one’s career (determined by income) following graduation, for students holding summer internships and for the control group of students who do not obtain competitive positions during their time in college.

For my data I will be using longitudinal educational studies that follow students through college and continue to document these participants as they enter the job market. First I will look at the correlation between school selectivity and holding summer internships. Next I will look at the correlation between summer internships and income following graduation. I do not currently have results but these, in combination with several controls including gender, age, and race, are the variables I am working with.

The limitations associated with my study will arise mostly from the datasets that are available. In order to obtain a random sampling, using existent datasets is much more effective for the purposes of this study than running a survey on a sample population and asking the questions I need, since this sample would be much smaller and rather selective instead of random. Unfortunately, this also means that I must make a lot of assumptions. In many of the datasets, there are variables that indicate jobs held during college, but often these variables are not very specific, and so I will be taking some liberty with determining whether these are to indicate internships or just regular non-competitive jobs. This will provide for some error. There will also exist, as often does, the problem of lurking variables and things I cannot control for. It is difficult to conclude whether it is the internship that causes the person to earn a higher income, or if the person that holds the internship will earn this higher income with or without this additional variable and it is simply that it is a type of person. Though the sampling is random, school choice is not. The participants in the studies were not randomly placed in selective and non-selective colleges, but rather made the choice of where they want to attend. This provides for sampling error. Despite these concerns, I do believe that I should be able to establish some viable correlation.


Comments:
Marlena,

Are you able to obtain the data from the College Student Survey or the National Survey of Student Engagement?

If so, I would encourage you to focus on the question "Does attending a more selective college make one more likely to participate in an internship?" I think that this is an interesting question because it provides insight into what might be different about attending selective colleges that make them worth attending. Also, those datasets include precise data on internships, so you don't have to worry about "manufacturing" them out of other questions.

This would be the dream approach. Get data from the CIRP freshman and college student surveys that have been linked together (I think that this is possible). Then, controlling for their high school grades, SAT scores, number of schools applied to, etc., regress whether or not they had an internship while in college on the selectivity of the school they attend (measured using the approach in either the Hoxby or Krueger Dale papers).

If you want to stick with the LR effects idea, I suggest that you slightly change focus from internships to "type/industry" of during school employment. That is, you want to see if attending a more select college affects the type or industry of the jobs one holds during college (particularly, if you can ID this, during the summer). Again, the basic approach is to, controlling for HS grades, SATs, etc., regress summer job type on college selectivity.

If you can pull off this second one, I am actually warming more to it. Some professions, like accounting, require students to do internships, so there may not be a selective college effect. What I think you are interested in really is whether or not attending a more selective college makes you more likely to obtain more "valuable" summer employment (and at Harvard this is synnonomous with internships, but that may not be the case everywhere). This approach might allow you to estimate the effect of selective colleges on this.
 
I like the idea, and going through the process myself I am definitly interested. I think the main problem with your regression will be that there is a large possibility for omitted variable bias even if you do control for all the factors that Bryce listed. Looking at the Malcolm Gladwell article on college admissions we come to realize that college admissions are now based on a lot more than just you GPA and SAT scores. There is a distinct social factor and a clear "ability" factor that differentiates kids who have the same GPA and SAT score. Students with better social skills and a higher "ability" are more likely to get into a selective insitution and also more likely to get prominent internships. In this case, its not the treatment affect of the school or even being part of the school itself (treatment effect aside) but the persons ability and social skills that got them the internship. I think you need an instrument to overcome this bias. Im not sure if there is data available on this, but I would suggest using students displaced from New Orleans following Katrina to more selective institutions. This should be uncorrelated with the error term as they did not show behaviours associated with getting into an Ivy but had the affect of being part of the Ivy. The only problem with this is that the dataset is probably pretty small and Im not really sure how liberal institutions were with their career resources for visitors.
 
Post a Comment

Subscribe to Post Comments [Atom]





<< Home

This page is powered by Blogger. Isn't yours?

Subscribe to Posts [Atom]