Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Panel data: Difference-in-differences and PSM/NN matching for firms with different event dates

    Hi

    I want to test the impact of firms receiving investment on firm turnover. I have a panel data covering c50,000 firms and 10 (consecutive) years of data per firm.

    Investment is a dummy variable (=1 for firms who receive the investment (treated); 0 otherwise (control)). I also have a DealYear variable which shows the year in which each firm received the investment. Post is a dummy equaling 1 where Year>= DealYear for treated firms (ie. it is 0 before they receive investment and 1 after).

    I want to match on industry group (a time-static variable) and TotalAssets in the year directly before the DealYear. ie. if a firm received investment in 2006, I want to match on the industry group and TotalAssets of the treated firm and the control firm in 2005. Thus I am matching firms which, in each case, were similar in the year before the treated firm received investment.


    As such, there are two elements I am currently unsure of incorporating into my approach:


    1. The event date for each treated firm is different
    2. I want to match on TotalAssets in the pre-event year in each case, so this will also differ for each firm


    I am matching with replacement, ie. the same control can be used for more than one treated firm.

    I have attached a simplified dataex data example to illustrate.


    Code:
    * Example generated by -dataex-. To install: ssc install dataex
    clear
    input str10 FirmName float(Year InvestmentDummy IndustryGroup Turnover TotalAssets DealYear Post)
    "Alpha"      2001 1 44 1019 10 2005 0
    "Alpha"      2002 1 44 1918  8 2005 0
    "Alpha"      2003 1 44 1111  8 2005 0
    "Alpha"      2004 1 44 1123  9 2005 0
    "Alpha"      2005 1 44 1234  7 2005 1
    "Alpha"      2006 1 44 1555 12 2005 1
    "Alpha"      2007 1 44 2333 13 2005 1
    "Alpha"      2008 1 44 2112 15 2005 1
    "Alpha"      2009 1 44 2121 21 2005 1
    "Alpha"      2010 1 44 1222 17 2005 1
    "Lok"        2001 0 12   33 14    . .
    "Lok"        2002 0 12   44 12    . .
    "Lok"        2003 0 12   59 12    . .
    "Lok"        2004 0 12   34 15    . .
    "Lok"        2005 0 12   32 19    . .
    "Lok"        2006 0 12   47 19    . .
    "Lok"        2007 0 12   45 11    . .
    "Lok"        2008 0 12   32 14    . .
    "Lok"        2009 0 12   11  9    . .
    "Lok"        2010 0 12   12 12    . .
    "Cooler Ltd" 2001 1 88  333 12 2008 0
    "Cooler Ltd" 2002 1 88  332 13 2008 0
    "Cooler Ltd" 2003 1 88  356 14 2008 0
    "Cooler Ltd" 2004 1 88  368 17 2008 0
    "Cooler Ltd" 2005 1 88  311 28 2008 0
    "Cooler Ltd" 2006 1 88  395 27 2008 0
    "Cooler Ltd" 2007 1 88  346 14 2008 0
    "Cooler Ltd" 2008 1 88  213 14 2008 1
    "Cooler Ltd" 2009 1 88  217  9 2008 1
    "Cooler Ltd" 2010 1 88  298 12 2008 1
    end



    I searched through the forum to try to find a solution but I was only able to find this similar post: https://www.statalist.org/forums/for...=1591791272650 but there was no definite solution.

    I hope the above is as clear and as detailed as possible.


    Any advice is appreciated,
    Paul

  • #2
    Hi Paul,

    Did you already find a solution for this problem?
    I face similar issues in my dataset.

    Best,
    Roger

    Comment

    Working...
    X