For my master thesis, I am studying the effect of input trade liberalizaion on Research and development decision depending on firms' financial constraints in India. I have a firm level database from 1991 to 1997 (before the asian crisis).
At a first step, I run a baseline regression with Research and development dummy as a dependant variable and input tariff, output tariff, interacted term between input tariff and importer status and some control variables. This works fine. A decrease in input tariff increases the probability to R&D. I used a fxed effect logit model and a linear probability one.
At a second step I use Rajan and Zingales's external financial dependance. I split my sample into high constrained firms and low constrained firms and I run the exact same model for the two groups (i.e R&D dummy as dependant variable and same explicative variables). I find that the highest constrained do more research and development. I compare the two coefficients in front of input tariff variables.
At a thrid step, I use an endogenous financial variable constructed from my database = leverage ratio. This is a robustness check for rajan and zingales' method When I slipt my sample into two groups high constrained and low constrained using leverage ratio, I find the opposite, the highest constrained are doing less research and development.
Do you have any advice for me to improve my work?