I am doing some work on the effect of formal/informal finance on firm performance sub-Saharan Africa. My main independent variable (PF) is a dummy variable. PF equals 1 if the firm finances working capital with formal sources (bank/microfinance) and 0 if they use informal sources.
The literature I have reviewed highlight a selection problem in the firm/owner's decision to go in for bank finance (Ayyagari, M., Demirgüç-Kunt, A., & Maksimovic, V. (2010). Formal versus Informal Finance : Evidence from China. The Review of Financial Studies, 23(8), 3048–3097). They cater for this problem by specifying a Heckman two step selection model. The first step is:
PF = B0 + B1Collateral + Other control variables + u, where Collateral is the identifying variable in the selection equation, since it affects the ability to access bank finance.
The second step is:
Performance = B0 + B1PF + Other control variables + u.
My problem is with how to implement this in Stata since most of the resources I find online seem to cater of selection bias in the dependent variable of the outcome equation.
This is what I try to do in Stata
Code:
heckman lnsales past_fin exper own_educ i.femlrgst_rec size i.agecat i.sector i.country, select(past_fin = iv_guarantee own_educ i.femlrgst_rec size i.agecat i.sector i.country) twostep
The selection equation however works fine (Attached in statalist2.png)
Any help would be greatly appreciated.
0 Response to Sample selection bias in independent variables
Post a Comment