HI
I hope you're doing well.
I'm writing to explain a limitation I encountered while performing propensity score matching using psmatch2, and how I addressed it in the current version of the analysis.
Issue Summary
In the original setup, I intended to match treated and control firms while accounting for both industry (ff_12) and year. These variables were included in the propensity score model via i.year i.ff_12 as follows:
However, upon examining the matched pairs, I discovered that psmatch2 does not enforce exact matching on year or industry. Including i.year and i.ff_12 in the command simply adjusts the estimation of the propensity score but does not restrict matches to occur within the same year or industry. As a result, some treated firms were matched with control firms from different years or industries, which could introduce biases into the analysis.
I hope you're doing well.
I'm writing to explain a limitation I encountered while performing propensity score matching using psmatch2, and how I addressed it in the current version of the analysis.
Issue Summary
In the original setup, I intended to match treated and control firms while accounting for both industry (ff_12) and year. These variables were included in the propensity score model via i.year i.ff_12 as follows:
HTML Code:
psmatch2 hard_final_Exact_new Fund_Status_w ROA_w Firm_Size_w i.year i.ff_12, neighbor(1) noreplacement
However, upon examining the matched pairs, I discovered that psmatch2 does not enforce exact matching on year or industry. Including i.year and i.ff_12 in the command simply adjusts the estimation of the propensity score but does not restrict matches to occur within the same year or industry. As a result, some treated firms were matched with control firms from different years or industries, which could introduce biases into the analysis.
