Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Modeling a question with unusual panel data

From   Mike McDonald <>
Subject   st: Modeling a question with unusual panel data
Date   Fri, 14 Dec 2012 12:05:27 -0500

Hi Everyone:

I have a research question that I'm having trouble figuring out how to
model. Any advice you all have is greatly appreciated.

My primary independent variable of interest is a measurement of
intellectual property protection -- it is panel data at the country
level, although it is not continuous. We have a measurement on the
variable every 5 years from 1960-2005 across 122 countries.

For the most part, each country has increased over time the level of
IP protection measured by this variable. For some countries, there is
a small increase every 5 years (each time we have a measurement). For
other countries, the variable jumps significantly at just one point in

I want to know the impact a change in IP protection has on a given
dependent variable.

It doesn't seem smart to treat this like normal panel data and run FE
or RE models on it: on the one hand because I only have data every 5
years, and on the other hand because most countries don't have
variation across all temporal periods but instead have a single (or a
few) big jumps. Plus, there is serial autocorrelation on the dependent
variables, so that adds a wrinkle, and I certainly don't want to get
into Bond estimators.

What I really want to model is this question: When the IV changes,
what is the impact on the DV in the following decade, compared to the
decade preceding the change. Or something like that.

Does anyone have a recommendation for how to model this?

Thanks for your time!
Mike McDonald
University of Maryland
*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index