I am conducting research in which I want to investigate the effect of tax incentives on research and development intensity in a group of firms. I have access to the data of a survey that:
- It's only for one year. (It has a cross-sectional nature).
- The companies were not chosen randomly, but those companies that conduct research and development.
- It includes nearly 3000 companies, about 20-30% of which have been exposed to tax incentives.(As I know, there are about 1,500 companies that have done research and development but not cooperated with the questioners)
- My dependent variable, which is the share of research and development expenses to the total expenses of the company, has a number between 0 and 1.
I'm having a little trouble choosing the right econometric method for this research.
#microeconometrics
For your research on the effect of tax incentives on R&D intensity in a group of firms, consider the following econometric method choices:
Logistic Regression with a Continuous Dependent Variable,
Since your dependent variable is continuous and bounded between 0 and 1, you could use a logistic regression model with the dependent variable transformed into a binary outcome. You can create a binary variable indicating whether a firm's R&D intensity exceeds a certain threshold (e.g., 0.5). Then, use logistic regression to model the probability of a firm exceeding this threshold based on tax incentives and other independent variables.
Beta Regression
Given that your dependent variable represents proportions, you might consider a Beta regression. Beta regression is specifically designed for modeling data bounded between 0 and 1. It can handle continuous dependent variables in this range effectively.
Propensity Score Matching (PSM):
Since your sample isn't randomly chosen and you have treated and control groups (firms with and without tax incentives), PSM could help address selection bias. PSM identifies firms with similar characteristics and propensity scores, allowing you to compare the effect of tax incentives on R&D intensity while mitigating selection bias.
Remember to control for potential confounding variables, such as firm size, industry, or other factors that could influence R&D intensity. Additionally, consider robustness checks and sensitivity analyses to ensure the robustness of your findings.
Thank you
consider a logit model for grouped data developed as an aggregate of people's choices so there is a consistent link to preferences
Dear Sadam Jamaldin and V. Kerry Smith
Thank you very much for your answers and tips.
- Badges
- Science topic
- Similar topics
- Mathematics
- Algebra
- Matrix